JP2010278625A - Network quality measurement system and method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve precision of quality measurement of a network in an inexpensive facility using a general-purpose computer terminal, such as a personal computer. <P>SOLUTION: A PC for measurement captures a communication packet by a counter value acquisition part 32b in a counter value acquisition wrapper program processing part 32, obtains processing load amount (counter value) of an OS when the communication packet is captured by the counter value acquisition part 32b, and a quality calculation part 33 calculates the quality of the network to be measured by using only the communication packet captured when the processing load amount (counter value) of the OS obtained is equal to or smaller than a predetermined threshold. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a technique for measuring a quality of a network by capturing a communication packet in a packet communication network such as IP (Internet Protocol) communication, and in particular, an inexpensive general-purpose computer device such as a PC (personal computer). The present invention relates to a technique suitable for measuring the quality of a used network with high accuracy.

  As a technique for measuring network quality, (1) a technique using an inexpensive general-purpose PC (personal computer; personal computer), and (2) an expensive dedicated hardware measuring instrument as described in Non-Patent Document 1, for example. There is technology to use.

  The technique using an expensive dedicated hardware measuring instrument as in (2) has higher accuracy, but is expensive, so that a measuring technique using an inexpensive personal computer as in (1) is often used.

  A major problem when using the personal computer of (1) is that the time accuracy of packet transmission / reception is not always high. The main reason for this is that it is not configured with dedicated hardware and software, so the time stamping process for packets by the PC's OS is delayed due to competition with processes other than the PC's internal measurement processing. The accuracy of the stamp is deteriorated.

[MD1230] Anritsu MD1230 [Search on May 22, 2009], Internet <URL: http: // www anritsu. co. jp / J / Products / tm / list. aspx? sID = 193 # 1481>.

  The problem to be solved is that, in the conventional network quality measurement technology using a general-purpose computer terminal such as a personal computer, the time stamp of the packet by the OS of the personal computer due to competition with processes other than the measurement processing inside the terminal. The giving process is delayed, and the accuracy of the time stamp deteriorates.

  An object of the present invention is to solve these problems of the prior art and improve the accuracy of network quality measurement in inexpensive equipment using a general-purpose computer terminal such as a personal computer.

  In order to achieve the above object, according to the present invention, for example, when measuring network quality from a communication packet captured using a personal computer (PC terminal), the operation state in the PC terminal is combined with the capture processing of the communication packet. If the processing load of the PC terminal is high when measuring the network quality from the captured communication packet, compare the observation result with the processing load amount of the PC terminal in normal times. It is characterized by improving the accuracy of quality measurement by removing it from the target.

  According to the present invention, the measurement value measured when the processing load of the PC terminal is high is rejected because it is determined that the probability of including an error caused by the measurement PC itself is high, and the measurement accuracy is effectively improved. Can be achieved.

It is a block diagram which shows the structural example of the network to which the network quality measurement system which concerns on this invention is applied. It is a block diagram which shows the structural example of the network applied to the setting of the threshold value in the network quality measurement system which concerns on this invention. It is a block diagram which shows the internal structural example of the network quality measurement system which concerns on this invention. It is a flowchart which shows the processing operation example of the network quality measurement system which concerns on this invention.

  Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings. FIG. 1 shows a configuration example of a network to which a network quality measurement system according to the present invention is applied, and a measurement PC 11 and a measurement PC 12 are connected via a measurement target network 13 such as the Internet.

  In FIG. 2, the measurement PC 11 and the measurement PC 12 in FIG. 1 are connected via the switching hub 23 via the network interface 11a and the network interface 12a. In such a connection form, the measurement PC 11 And the measurement PC 12 calculate and set the threshold value according to the present invention by the threshold value setting processing units 11b and 12b provided respectively.

  Also, in FIG. 3, the measurement PC 11 and measurement PC 12 in FIG. 1, which are characteristic of the present invention, are shown together with the processing form by the measurement program processing unit (described as “measurement program” in the figure) 31 in a conventional general-purpose PC. The measurement PC 11 and the measurement PC 12 are each a counter value acquisition wrapper program processing unit (“counter value acquisition wrapper” in the figure) as means for executing computer processing by a program operating on the operating system. The counter value acquisition wrapper program processing unit 32 includes a measurement program processing unit (described as “measurement program” in the figure) 32 a and a counter value acquisition unit (in the figure). (Described as “counter value acquisition”) 32b.

  The measurement PC 11 and the measurement PC 12 in FIG. 1 equipped with the network quality measurement system having the configuration shown in FIG. 3 capture the communication packet by the measurement program processing unit 32a in the counter value acquisition wrapper program processing unit 32, and The processing load amount (counter value) of the operating system (OS) at the time of capturing the communication packet is acquired by the counter value acquisition unit 32b in the counter value acquisition wrapper program processing unit 32, and acquired by the quality calculation unit 33 The quality of the network 13 is calculated using only the communication packet captured when the processing load (counter value) of the operating system is equal to or less than a predetermined threshold.

  As described above, in the counter value acquisition wrapper program processing unit 32 of this example, the counter value acquisition unit 32b obtains a counter value indicating the internal state of the operating system at the time of capturing the communication packet, and the counter value is processed by the operating system. Calculate as load.

  Further, as shown in FIG. 2, the measurement PC 11 and the measurement PC 12 in FIG. 1 measure the counter value indicating the internal state of the operating system a plurality of times by the threshold setting processing units 12a and 12b. A threshold value is calculated using the counter value.

  At this time, the threshold value setting processing units 12a and 12b use the counter value measured at the time of capturing the communication packet via the switching hub 23 for calculating the threshold value.

  Hereinafter, the details of the technique relating to the network quality measurement system of this example having such a configuration will be described.

  When a recent high-performance PC is used for measuring network quality, the phenomenon of accuracy degradation that occurs when a conventional PC is used does not occur frequently but occurs in a limited manner.

  In network quality measurement, in many cases, measurement is performed multiple times instead of “one measurement and finished”, and evaluation is performed with statistical values (average, maximum value, 99.9% value, etc.).

  For example, ITU-T Y.M. 1540 Recommendation [Y. 1540] defines [99.9% of packet transfer one-way delay value−minimum value] as a delay fluctuation index. To continuously monitor the delay fluctuation, the value of 99.9% of one-way delay is defined. Must be continuously measured several times.

  Of these multiple measurements, if the abnormal values due to process processing conflicts, etc., which was a problem when measuring network quality using a general-purpose PC, can be excluded and aggregated, Accuracy is also improved.

  However, even if an abnormally large delay time is observed in the measurement using a general-purpose PC, the abnormal observation value is “abnormal value due to processing delay of measurement PC” from the network quality measurement result itself. It is not possible to determine whether or not it is “a correct measurement result due to a large network transfer delay time”.

  In this example, focusing on the fact that some OSs of general-purpose PCs have counters for monitoring various internal states, the value of this counter is recorded before and after each quality measurement. It is determined whether or not the OS load was high at the time of measurement.

  In other words, the measurement value measured in the time zone in which the value of this counter changes greatly from the normal time has a high probability of including an error due to the measurement PC itself. The measurement accuracy is effectively improved by rejecting

  As for the “threshold value”, since it is only necessary to compare with the normal time, some standard may be determined as appropriate.

  Hereinafter, an example using FreeBSD (registered trademark) as an example of an OS used for measurement will be described.

  As shown in FIG. 1, consider a case in which the quality of the network 13 is measured by transmitting and receiving test packets between the measurement PC 11 and the measurement PC 12 which are personal computers for measurement.

  In FreeBSD (registered trademark), processing at the time of packet reception is in principle directly processed in real time without being placed in the processing queue. However, when the direct processing is not in time for some reason, it is put in the queue. Yes. This leads to a processing delay that causes a time stamping error in a conventional general-purpose PC.

  In FreeBSD (registered trademark), the processing amount of the packet processing queue of the network stack is recorded by a counter inside the kernel “net.isr.queued”. This can be acquired with the “sysctl” command.

  A command execution example (# is a command input prompt) for obtaining the counter value in this way is shown below.

#Sysctl net. isr. queued
23989

  Therefore, the command is executed immediately before and immediately after the measurement is performed once, and the increment of the output value (immediate value−immediate value) is calculated.

  In order to describe a specific example in which this technology is actually used, a case where a measurement program having a command name “measurement_command” is used will be described.

  Any measurement command may be used, and a commonly used measurement program (iperf [IPERF], opening [OWAMP: One-Way Active Measurement Protocol], etc.) can be used, and the type of application is not limited.

  Here, it is described with the name “measurement_command”, but the description in this example can be replaced by other applications as it is.

  Although there is no need to modify the measurement application (measurement_command), the technology of this example needs to know the start time and end time in order to grasp the state of the system at the time of executing it. is there.

  A technique of periodically recording a counter value and collating it with the time of a measurement result later, or a technique of starting a measurement program with a wrapper program and recording a counter value before and after that as shown in FIG. There is also.

  Hereinafter, the latter shown in FIG. 3 (activation of the measurement program by the wrapper program) will be described as an example. Here, an example in which the wrapper program is implemented by a UNIX (trademark) system zsh shell script for processing will be described.

  Here, the name of the wrapper program is “record_counter sh”. The contents of “record_counter sh” are shown below.

#! / Usr / bin / env zsh
counter_before = $ (sysctl net.isr.queued) # Save the previous value as a variable measurement_command # Measurement execution (command name depends on the measurement tool used)
counter_after = $ (sysctl net.isr.queued) #Save the value immediately after as a variable counter_incr = $ ((counter_after-counter_before)) #Calculate the difference echo $ counter_incr #Output the difference

  At the time of measurement, instead of directly executing “measurement_command”, by executing “record_counter.sh” (“measurement_command” is indirectly executed in the measurement), the difference in counter value at the time of measurement is performed. Can be output simultaneously.

  In bidirectional measurement, as shown in FIGS. 1 and 2, the same counter value is recorded before and after the measurement time on the remote side.

  If this is repeated a plurality of times, the following data is obtained.

Measurement result (value), local instrument counter increment, telemeter counter increment Measurement result (value), local instrument counter increment, telemeter counter increment Measurement result (value), local instrument counter increment Remote counter counter increment Measurement result (value) Local counter counter increment Telemeter counter increment Measurement result (value) Local instrument counter increment Telemeter counter increment Measurement result (value) Local counter counter increment Telemeter counter increment Measurement result (value) Local counter counter increment Telemeter counter increment
...

  Here, an example of how to determine the “prior threshold” is shown.

  Before actually performing measurement in the measurement target network 13 shown in FIG. 1, two test PCs are directly connected via the switch hub 23 in advance in a test environment as shown in FIG. Do not pass through), conduct an evaluation test.

  Since the counter increment depends on the model used for measurement, the measurement method used, and the packet transmission / reception pattern, the “preliminary evaluation test” using the same measurement method / packet transmission / reception pattern as the measurement planned for the network 13 to be measured. Is performed multiple times to obtain the counter increment value.

  The threshold is determined from the “counter increment value distribution” at that time.

  The method of determining the threshold value is arbitrary, but here, the distribution at normal time is almost uniform and it is desired to detect an abnormal value that occurs very rarely. For example, (99% value of distribution + 1) A threshold value of “abnormal” value exceeding “time value (≈99% value of distribution)” can be determined.

  When the prior threshold value is determined in this way, the following processing is performed on the measurement data string.

  First, one line of measurement data is acquired, and it is determined whether the local counter increment is equal to or less than a reference value (Yes) or not (No). If not (No), the data is rejected, and Yes If there is, then it is determined whether the remote counter increment is equal to or less than a reference value (Yes) or not (No). If not (No), the data is rejected, and if Yes, the measurement result Since (value) is adopted as a correct measurement result, it is recorded in a storage device (not shown), and the process for the next data is repeated.

  A set of measurement results that are not rejected by such processing is a more accurate measurement result because data with a large measurement error is rejected.

  Hereinafter, a processing procedure example of the network quality measurement method according to the present invention will be described with reference to FIG. In this example, it is assumed that N capture processes are performed.

  Before capturing the communication packet (step S402), the value Lb of the operation state counter (net.isr.queued) in the measurement PCs 11 and 12 is acquired and stored in the memory (1) 41 (step S401). After capturing the communication packet and storing the data in the memory (1) 41, the value La of the operation state counter (net.isr.queued) in the measurement PCs 11 and 12 is acquired and stored in the memory (1) 41. Store (step S403).

  After the above process is repeated N times (steps S404 and S405), the following process is repeated N times using the N pieces of data stored in the memory (1) 41 (steps S406 to S411).

  First, the counter values Lb and La stored in the memory (1) 41 are read for the first time, and the increment ΔL of the counter is calculated from La−Lb (step S407). Then, it is determined whether or not the calculated increment ΔL is equal to or less than the threshold (reference value Lo) previously obtained by the threshold setting processing units 11b and 12b in FIG. 2 (Yes) or not (No) (step S408). .

  If NO (No), the first capture result data is rejected (step S409). If YES, the capture result data is stored in the memory (2) 42 (step S410).

  After the processes of steps S407 to S410 are repeated N times (step S411), the quality of the network (13) is calculated using the capture result data stored in the memory (2) 42 (step S412).

  As described above with reference to FIGS. 1 to 4, in the network quality measurement system and method of this example, when measuring the network quality of the packet communication network 13 by the general-purpose PC terminal, the measurement PCs 11 and 12 Along with the quality measurement of the network 13, a counter representing the internal state of the installed OS is measured, and it is determined whether or not the load on the OS is high compared with a predetermined threshold value. When the measured value of the counter is higher than the threshold value, the measured network quality is excluded from the measurement target, and the network quality is calculated using only the remaining measurement result.

  Note that the threshold is determined by performing counter measurement a plurality of times by transmitting and receiving packets at two terminals directly connected via the switching hub 23, and using the statistical value as a threshold.

  As described above, according to this example, the measurement value measured when the processing load on the PC terminal is high is judged to have a high probability of including an error caused by the measurement PC itself, and is effectively rejected. It is possible to improve the measurement accuracy.

  In addition, this invention is not limited to the example demonstrated using FIGS. 1-4, In the range which does not deviate from the summary, various changes are possible. For example, in this example, FreeBSD (registered trademark) is described as an example of the OS, but other types used for general-purpose PCs such as Linux (registered trademark), NetBSD (registered trademark), and BSD / OS (trademark). It is also applicable when using any OS.

  11, 12: Measurement PC, 11a, 12a: Network interface, 11b, 12b: Threshold setting processing unit, 13: Measurement target network, 23: Switching hub, 31: Measurement program processing unit, 32: Counter value acquisition wrapper program Processing unit, 32a: measurement program processing unit, 32b: counter value acquisition unit, 33: quality calculation unit.

Claims (6)

A network quality measurement system that calculates network quality of a packet communication network from captured communication packets by computer processing by a program operating on an operating system,
As a means of performing programmed computer processing,
A first means for acquiring a processing load of the operating system at the time of capturing the communication packet;
And a second means for calculating the network quality using a communication packet captured when the acquired processing load of the operating system is equal to or less than a predetermined threshold. The network quality measurement system according to claim 1,
The first means includes
A network quality measurement system, comprising: obtaining a counter value indicating an internal state of the operating system at the time of capturing the communication packet, and calculating the counter value as a processing load amount of the operating system. A network quality measurement system according to claim 1 or 2, wherein
As a means of performing programmed computer processing,
A network quality measurement system comprising a third means for measuring a counter value indicating an internal state of the operating system a plurality of times and calculating the threshold value using the counter value a plurality of times. The network quality measurement system according to claim 3,
The third means is
A network quality measurement system using the counter value measured at the time of capturing a communication packet via a switching hub for calculating the threshold value. A network quality measurement method for a system that calculates a network quality of a packet communication network from a captured communication packet by computer processing by a program operating on an operating system,
As a procedure executed by the programmed computer,
5. A network quality measuring method comprising a processing procedure executed by each means in the network quality measuring system according to claim 1.   The program for functioning a computer as each means in the network quality measurement system in any one of Claims 1-4.

Images