JP2000069088A - Traffic evaluating method for packet exchange network, recording medium recording program for poviding the same and traffic evaluating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To evaluate a traffic load concerning a path to an object host or respective links composing the path while considering directivity as well and further to independently measure the bands of respective links in respective directions. SOLUTION: A probe packet 10, that time T1 just before transmission from a transmission host to a recipient is recorded in a time stamp header 12, is transmitted and when this packet is received, the recipient generates and transmits a response packet 20 toward a transmitter. Relating to this response packet 20, three kinds of time of time T2 for the recipient to receive the probe packet 10, time T3 to transmit the response packet 20 and said time T1 recorded in the probe packet 10 are recorded in its time stamp header 22. The response packet 20 recording these three kinds of time is returned to the transmission source.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a transmitting host as an end terminal, which transmits a probe packet having a time stamp (time stamp) header to a receiver. The time is written in the packet, the packet is sent back as a response packet to the sending host, and the sending host counts the above-mentioned time stamp header information to evaluate the traffic load of any link or path in the packet switching network. Method,
The present invention relates to a recording medium on which a program for realizing this method is recorded and a traffic evaluation device.

[0002]

2. Description of the Related Art As a conventional path characteristic evaluation system using this kind of probe packet, a pathchar which is a tool generally used on the Internet is used.
It has been known. In this system, the Internet sends multiple probe packets to all routers along the path from the sending host to the destination host, receives response packets corresponding to them, and round-trip delay to each node. By summing the time at the sending host, the waiting time in the queue for one hop in the forward direction is evaluated for all the links between the routers. In this case, the relay node that has received the probe packet generates a response packet to be transmitted to the transmitting host at the same time as receiving the probe packet. The data part of this response packet is the first 64 of the received probe packet.
Since the data for the byte is copied, the size of the response packet is constant. In this regard,
For example, Okada et al., "The Contents of the Internet Measurement Network Can Be Seen" (bit, August 1998,
The description of Kyoritsu Shuppan can be referred to.

[0003]

The purpose of a traffic load evaluation system based on probe packets is to transmit a probe packet from its own end terminal and receive a response packet to measure the delay inside the network. When performing traffic control such as selecting the optimal target host, etc., based on the aggregation result, grasping the path extended from the end terminal to the target host or the load state and bandwidth of each link constituting the path, and selecting the optimal target host And
It is useful for network management.

A conventional path characteristic evaluation system based on such delay time measurement is based on the following concept. That is, the round trip time Tn, s (s: packet length) from the measurement host to the n-th router is represented by a transfer delay proportional to s, a fixed amount depending on the link distance, and the like, and a traffic load (queue of the router). And a variation depending on the waiting time). If a sufficiently large number of samples are taken for each router, it is highly probable that at least one of them has zero latency at each router, so the minimum value does not include the variation in delay. Can be The deviation qn 'from the minimum value such as the average value of Tn, s can be considered to be the sum of the fluctuations at each node from the sending host to the n-th router. Therefore, the variation qn of the round trip delay of the nth hop can be calculated from qn = qn'-qn-1 '.

Since the above-mentioned qn 'is the variation of the round trip delay, the sum of the forward and return variations is evaluated. In addition, the variation qn of the delay of one hop is represented by two queues Qn-1 (→) in FIG. 1 (indicating the forward direction, including in the figure, the same applies hereinafter).
And Qn (←) (the same applies to the following directions, including the direction in the figure). These two queues are forward queue and reverse queue, respectively, and their traffic load can be evaluated only as a sum, and their directivity cannot be evaluated.

Further, regarding the above-mentioned minimum value of Tn, s,
Consider changing the packet length s. If a sufficient number of samples are taken, the probability that the minimum value of Tn, s for each s is composed of a fixed delay and a transfer delay proportional to the packet size is high. _Where bi is the band of each link, (slope of min Tn, s) = Σ _{(i =} 1 to _2n) (1 / bi). Here, it is assumed that the size of the response packet is the same as that of the probe packet. Also, Σ
_{(i =} 1 to 2n ₎ means that the sum is obtained from i = 1 to i = 2n. In the above-mentioned pathchar,
Since the size of the response packet is constant, (slope of min Tn, s) = Σ _{(i =} 1 to _n) (1 / bi).

Therefore, (slope of min Tn, s) − (min Tn
1bn for the nth hop
(→) + 1 / bn (←) can be calculated. Where bn
(→) and bn (←) are bands in the forward and reverse directions, respectively. In pathchar, only 1 / bn (→) can be measured. Also, even if the size of the response packet is equal to the size of the probe packet, bn (→) and bn
(←) cannot be separated.

An object of the present invention is to solve the above-mentioned problems in the prior art and to send a probe packet to a relay node or a target host, thereby forming a path to the target host or the path to the target host. For each link that is configured, the traffic load is evaluated in consideration of the direction (difference in the degree of congestion between going and returning), and the bandwidth of each link (the forward direction and the reverse direction are not necessarily equal).
Is to be independently measurable in each direction.

[0009]

In order to achieve the above object, in the traffic evaluation method and apparatus according to the present invention, the transmission host records the time T1 immediately before transmission to the receiver in the time stamp header 12. Probe packet 1
0, and the receiver generates and transmits a response packet 20 directed to the sender at the same time as receiving this. The above-described response packet 20 includes the time T2 when the probe packet is received by the probe packet receiver, the time T3 when the response packet 20 is transmitted, and the time T1 recorded in the probe packet 10 as the time stamp header. 22
And the data part 23 is a copy of the data part 13 of the probe packet 10. The response packet 20 in which these three times are recorded is returned to the transmission source. In this case, the clock of the sending host and the clock of the receiver do not need to be synchronized.

[0010] The transmitting host records the time T4 at the moment when the response packet is received, whereby Dn (→) =
Two quantities are obtained: T2−T1, Dn (←) = T4−T3. This amount is the one-way delay time from the sending host to the receiver plus the clock skew of both. On the other hand, since it is possible to assume that the clock deviation is constant during the measurement time, the variation of D (→) and D (←) is different in the order of the path from the transmission host to the receiver. It is equal to the one-way delay variation in the opposite direction. Therefore, if these two quantities are separately tabulated, it is possible to independently measure forward and backward fluctuations. Also, the part of the delay that depends on the packet length is similarly not affected by the clock shift.

When evaluating the forward traffic load of a path, Dn (→) is measured a plurality of times, and for example, the deviation qn ′ (→) of the average value from the minimum value is used to evaluate the forward traffic load. Characterize traffic load. Also, qn (→) = q
The traffic load of the nth hop in the forward direction is characterized by n ′ (→) −qn−1 ′ (→). Further, the forward band of the nth hop is obtained from the result of measuring Dn (→) by changing the packet size s. Dn (→)
Is considered to be a linear expression of s, the reciprocal of the band in the forward direction of the n-th hop is obtained from (slope of min Dn (→)) − (slope of min Dn−1 (→)). It can be calculated.

Conventional techniques, for example, the above-mentioned pathch
In er, the receiver that has received the probe packet does not add any information of itself, only sends back a response packet, and has only a round-trip evaluation, so that it cannot distinguish the direction of the traffic load. On the contrary,
According to the present invention, the receiver embeds the time of its own clock in the response packet and sends it back, so that it is possible to separately evaluate the directionality of the traffic load.
Further, in the conventional technology, there is a problem that the measurement results of the bandwidths of the two links (forward and return) of each hop cannot be separated or only the bandwidth of the forward link can be measured. Can be measured.

A program for realizing the traffic evaluation method according to the present invention can be distributed as a product in a form in which the program is recorded on a recording medium, but the protection scope of the present invention is such a product. Needless to say, it extends to.

[0014]

Embodiments of the present invention will be described below in detail with reference to preferred embodiments shown in the drawings. FIG.
FIG. 2A shows the format of the probe packet 10 shown in FIG.
(b) shows the format of the response packet 20. At the beginning of the packet, basic headers (eg, IP headers) 11 and 21 for packet exchange are placed, followed by a time stamp 3
Time stamp headers 12 and 22 each having a size corresponding to the time stamp header are attached. After that, the data portions 13 and 23 of the packet follow.

The sending host embeds the time immediately before sending the probe packet (the sending time of the sending host) in the first field 12a of the time stamp header 11 of the probe packet 10, and sends it out thereafter. Immediately after receiving the probe packet 10, the receiver generates a response packet 20, and stores its own time stamp 22 a and time stamp 22 in the time stamp header 22.
b is embedded and then transmitted.

FIG. 3 shows an example of the configuration of an apparatus for performing traffic control based on the data on the directionality of the traffic load evaluated as described above. A client terminal 40 receiving a service from the information server 30 is connected to two networks A and B via a traffic evaluation device 50 according to the present invention. The service provided by the information server 30 performs bulk transfer of data such as TCP, transfers a large packet including information to the client terminal 40, and the client terminal 40 transmits a small ACK to the packet. Packets are sent back sequentially.

The traffic evaluator 50 is designed for use in cases where the bandwidth in the direction in which the data packet is transmitted (direction from the information server to the client: the thick arrow in the figure) is large or the traffic load in that direction is light. Control traffic to select a network.

Here, the traffic evaluation device 50 evaluates the traffic load of the two paths via the respective networks A and B in both directions, and based on a certain range of data up to the past, Is determined as to which path to use. In this determination, first, a physical band in a direction in which a data packet is transmitted is measured for each link, and a network having a larger bottleneck band is preferentially selected.

Further, the traffic load in the round trip is compared for the two paths, and if the difference exceeds a certain threshold, the network with the smaller traffic load in the round trip is selected. If the difference is less than the threshold, the network with the lighter traffic load in the direction in which the data packet is transmitted from the information server 30 is selected.

FIG. 4 shows a block diagram of the traffic evaluation device 50 described above. The probe packet transmission scheduling unit 1 determines a time interval for transmitting probe packets and the number of packets, and controls the probe packet generation unit 2a. The probe packet generator 2a sends out a probe packet according to a command from the scheduling unit 1. The response packet receiving unit 2b receives the response packet and attaches a time stamp at the time of reception. Further, the measurement data storage unit 3 acquires the time stamp information from the response packet receiving unit 2b, calculates the delay time, and stores it in the memory.

The band evaluation unit 4a and the traffic load evaluation unit 4b read out the delay time information recorded in the measurement data storage unit 3 and evaluate the band and the traffic load in each direction. Further, the probe packet scheduling unit 1 may reduce the evaluation error here.
Then, the parameters of the time interval and the number of packets are fed back. The routing table determination unit 5 determines a routing table for each occasion based on the evaluation results of the band evaluation unit 4a and the traffic load 4b. The route selection execution unit 6 executes packet routing based on the current routing table.

According to the above embodiment, the receiver embeds the time of its own clock in the response packet and sends it back. Therefore, it is possible to separately evaluate the directionality of the traffic load. become able to. Conventionally, the measurement results of the bandwidths of the two links (forward and return) of each hop could not be separated or only the bandwidth of the forward link could be measured, but these can be measured. It should be noted that the above embodiment is an example of the present invention, and it is needless to say that the present invention is not limited to this.

[0023]

As described above in detail, according to the present invention, the sending host acquires the time of the relay node or the receiving host inside the network, and grasps the one-way delay variation based on the time. Therefore, it is possible to calculate the traffic load that can be evaluated only in the conventional round-trip, and it has a remarkable effect that the evaluation of the bandwidth of each hop going and returning can be independently performed.

[Brief description of the drawings]

FIG. 1 is a diagram showing two links and two queues forming one hop.

FIG. 2 is a diagram illustrating an example of a packet format of a probe packet and a response packet.

FIG. 3 is a diagram illustrating a configuration example of a system that performs traffic control based on a measurement result of a bottleneck band in a direction of receiving data and a traffic load.

FIG. 4 is a block diagram of an apparatus according to an embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Probe packet transmission scheduling part 2a Probe packet generation part 2b Response packet reception part 3 Measurement data storage part 4a Band evaluation part 4b Traffic load evaluation part 5 Routing table determination part 6 Route selection execution part 30 Information server 40 Client terminal 50 Traffic evaluation device

Claims (5)

[Claims] 1. In a packet switching network having a plurality of relay nodes along a path from a sending host to a receiving host, the sending host sends a plurality of probe packets, each of which records a time immediately before sending, to all the relay nodes. Each receiving node receives the probe packet addressed to the own node and simultaneously records the transmission time stamp recorded in the probe packet by the transmitting host and the time when the own node receives the probe packet in a response packet to the probe packet. Then, the data part of the received probe packet is copied, and the time immediately before transmitting the generated response packet is recorded, and the response packet is returned to the transmission host. The transmission host returns the time stamp information of the received response packet. From each link between each node in the forward and reverse directions. Traffic evaluation method in a packet switched network, characterized by aggregating the click load and bandwidth. 2. In a packet switching network having a plurality of relay nodes along a path from a sending host to a receiving host, the sending host sends a probe packet recording the time immediately before sending to a single receiving host or a single A plurality of packets are sent to the relay node, and the receiver receives the probe packet of the self destination, and at the same time, in the response packet to the probe packet, the transmission time stamp recorded by the transmitting host in the probe packet and the receiver inserts the probe packet. The receiving time is recorded, the data part of the received probe packet is copied, and the time immediately before transmitting the generated response packet is recorded, and the response packet is returned to the transmitting host. From the time stamp information of the packet, the path between the sending host and the receiver in the forward and reverse directions Traffic evaluation method in a packet switching network, characterized in that aggregates the traffic load. 3. A recording medium storing a program for implementing the method for evaluating traffic in a packet-switched network according to claim 1. Means for scheduling probe packet transmission and generating and transmitting the probe packet; means for receiving a response packet to the probe packet and storing measurement data; means for evaluating a bandwidth and a traffic load from the measurement data. And a means for totalizing the traffic load of each link in the forward direction and the reverse direction for all links between the nodes and selecting a communication route. 5. A means for scheduling probe packet transmission and generating and transmitting a probe packet, a means for receiving a response packet to the probe packet, storing measurement data, and a means for evaluating a bandwidth and a traffic load from the measurement data. A traffic evaluation apparatus for a packet-switched network, comprising: means for counting the traffic load of a path between a transmitting host and a receiver in the forward and reverse directions and selecting a communication path.