Queueing system H2/M/1 for calculation of the characteristics of network traffic Текст научной статьи по специальности «Компьютерные и информационные науки»

QUEUEING SYSTEM H2/M/1 FOR CALCULATION OF THE CHARACTERISTICS OF NETWORK TRAFFIC

DOI 10.24411/2072-8735-2018-10215

Sergey V. Malakhov,

Povolzhskiy State University of Telecommunications and Informatics, Samara, Russia, malakhov-sv@psuti.ru

Ekaterina M. Mezentseva,

Povolzhskiy State University of Telecommunications and Informatics, Samara, Russia, katya-mem@psuti.ru

Keywords: network traffic, sample variance, coefficient of variation, asymmetry, initial moments, hyperexponential distribution.

In due to development and introducing new technologies, the increase in the transmitted data acquires increasing urgency problem of analysis network traffic. The solution to the problem of network traffic analysis can be separated into 3 independent subtasks: capturing, storage and the analysis itself. The following article describes software for capturing network traffic. It includes a preliminary analysis of the data and provides the ability to intercept packets, extracting the necessary parameters from them, such as time, packet length, flags, and others. The appendix presents the possibility of extracting this information from the files and analyzing the captured packets either in real time or by processing the data received from the file. After capturing the network traffic, due to the use of the developed software, the receipt and packet sending time sampling was sampled. Next, the obtained data was analyzed and such parameters as the first, second and third initial moments, sample variance, the factor of variation and asymmetry were calculated. Analysis of captured packets is the calculation of parameters: the first, second and third the initial moments, sample variance, the coefficient of variation and the asymmetry. These statistics provide information about the nature of the distribution of time intervals. For example, the coefficient of variation indicates the difference between traffic and the Poisson's flow and in conjunction with the asymmetry gives information about the exponent of tails distribution weight. As a result, these parameters will be able to get the value of the delay of the traffic. The program also allows you to compare theory with practical data. The moment characteristics of the distribution of time intervals between the received packets are calculated using the mathematical statistics formulas. These statistics allow us to judge the distribution of intervals. The distribution of intervals between traffic packets relates to heavy-tailed distributions. Considering the weight of the tail of the input distribution the queuing system H2/M/1, gives a delay many times greater than the classical model, even at low loads.

Information about authors:

Sergey V. Malakhov, Assistant Professor, Software and Management in technical Systems Department, Povolzhskiy State University of Telecommunications and Informatics, Samara, Russia

Ekaterina M. Mezentseva, Assistant Professor, Software and Management in technical Systems Department, Povolzhskiy State University of Telecommunications and Informatics, Samara, Russia

Для цитирования:

Малахов С.В., Мезенцева Е.М. Система массового обслуживания H2/M/1 для расчета характеристик сетевого трафика // T-Comm: Телекоммуникации и транспорт. 2019. Том 1. №12. С. 56-59.

For citation:

Malakhov S.V., Mezentseva E.M. (2019). Queueing system H2/M/1 for calculation of the characteristics of network traffic. T-Comm, vol. 13, no.1, pр. 56-59.

■

COMMUNICATIONS

Introduction

This system must provide the eapture of all traffic, and provide methods of data analysis. To capture the traffic using special utilities - sniffers. Sniffer - a software or hard ware-software device designed to intercept traffic. Sniffer can installed on the router or on the network terminal device.

After the capture of packets must be, choose what to do with them: handle in real time, or save to a file for further analysis.

Currently there are several programs for the operating system Microsoft Windows, and for Linux. Mainly they used for traffic capturc, but traffic analysis will be carried out manually. In addition, not all software allow you to record the necessary data in a file in a form suitable for further processing. [ 1 ]

Such programs allow user to browse entire traffic passing through the network in real time, switching the network card into promiscuous mode (promiscuous mode).

Therefore, we need the software that will provide the ability to intercept packets, showing the values of various fields in the packet, such as lime, length, (lags, etc.; decoding from raw digital form into a human-read able the format, the analysis of captured packets in real-time or processing data obtained from the file. Therefore the software which calculates necessary characteristics of a network traffic was realized.

Determination of the moments of the interval time

of incoming traffic and delay

After the capture of network traffic, must be sampling the lime of send and receive packets. Using the formula of mathematical statistics, torque characteristics the distribution of time intervals between the accepted packets. [2|

Analysis of captured packets is the calculation of parameters: the first (1), second (3) and third (6) the initial moments, sample variance (2), the coefficient of variation (4) and the asymmetry (5). These statistics provide information about the nature of the distribution of time intervals. For example, the coefficient of variation indicates the difference between traffic and the Pois-son's flow and in conjunction with the asymmetry gives information about the exponent of tails distribution weight. As a result, these parameters will be able to get the value of the delay of the traffic (7), The program also allows you to compare theory with practical data.

The average value of the interval between packets [3j: 1 N A1 ¿=o

where t k~ times receiving packets, N— number of intervals.

(1)

N to C = Ge/ï,

where a, = ,

A,=f3H>

££3

Delay is:

(2)

(3)

(4)

(5)

(6)

W=---L CO

y, M

where - negative root of the integral equation Lindley. to solve the spectral method, - intensity of input stream. [4]

Program must be a graphical view of all the above parameters as the charts and graphs. The software to help automate the above process, thereby saving time and effort of people involved in this problem. [5J

Statement of the problem

1. Create the conditions presented by scheme (fig. 1 ).

2, Capture the traffic passing through PSUTl server and analyze it.

The object and subject of research. The object of research -the cluster traffic PSUTI. The subject of research - Characteristics of network traffic PSUTI calculated using formulas of mathematical statistics.

Min-ored port

Fig. 1. Network

Experiment

Experimenting carried out to check the correctness of the program and calculation of traffic parameters. The object of this experiment was selected traffic PSUTI. For a log file of all packets passing through the network, we need to connect to the server and enable the network card promiscuous mode to receive all the network packets, even those that not addressed to it.

After connecting to the server, you must start the program, select the file, and network adapter. By clicking on "packet capture" starts capturing all network packets, extracting the information, and recording this information in selected file (fig. 2).

Traffic Analizer

Settings Charts Formutas Fite

OAPSUTTtfcd

File optioni Scm-C« •ddr»»

adapter

o&ctt* -Reatefc PC* GBE famtf Oy v

Time 000020

Number of captured packets 0

Capture Stop Capture

0

y:

S:

Traffic Filter Create Data Chain Convert Split

Fig. 2, Settings

T-Comm Vol.13. #1-2019

7TT

Y

T-Comm Tом 13. #1-2019

Y

References

Conclusion Intermediate values

Analyze the obtained data (fig. 7), Packets number 10 , First initial time 4,17*10"4 s, Second initial time l,58*10~6s, Third initial time 1,11*10® s, Dispersion 1,4*10"*, Coefficient of variation 2,84, Asymmetry 6,68. From fig. 3 shows that the average value of the interval between packets is not more than 14-14.5ms. Next, consider a delay. As seen from the chart delay, it may be up to 10 seconds, which certainly bad for the server, even though such packets arc not so much (<0.3%).

The obtained data demonstrate that the analyzed traffic considerably differs from Poisson distribution {coefficient of variation c > 1). The value of asymmetry j4s > 2 indicates that the distribution of time interv als between the traffic packcts refers to heavy-tailed distributions and matches the queuing system (QS) Hj/M/l. The dispersion also has a low enough value. Using a formula we calculate average delay of a network on this interval of time above. As seen from fig. 7 delay, computed by formulas and packets delay with feedback, differ only on 17ms (=0,11%).

1. Statistical analysis of network traffic [Electronic resource]. Access mode: http://pi.314159.ru (01.17.2018).

2. Analysis of the network traffic in real time: a review of applications, approaches and solutions | Electronic resource]. Access mode: http://www.vanderboot.ru (01.17.2018).

3. Tarasov V.N., Bakhareva N.F., Gorelov G.A., Malakhov S.V. (2014). Analysis of the incoming traffic at three moments in time intervals distribution. Information technologies, pp. 54-59.

4. Tarasov V.N., Malakhov S.V. (2015). Statistical data handling program of Wireshark analyzer and incoming traffic research. Proceedings of the Institute for System Programming, vol. 27(3), pp. 303-314.

5. Tarasov V.N., Bakhareva N.F., Kartashevskiy [.V., Malakhov S.V. (2016). Analysis of intervals between traffic packets on the SDN networks depending on the TCP window size. Problems of Infocommunications Science and Technology (PIC S&T), pp. 15-17,

СИСТЕМА МАССОВОГО ОБСЛУЖИВАНИЯ Н2/М/! ДЛЯ РАСЧЕТА ХАРАКТЕРИСТИК СЕТЕВОГО ТРАФИКА

Малахов Сергей Валерьевич, Поволжский государственный университет телекоммуникаций и информатики,

г. Самара, Россия, malakhov-sv@psuti.ru

Мезенцева Екатерина Михайловна, Поволжский государственный университет телекоммуникаций и информатики,

г. Самара, Россия, katya-mem@psuti.ru

Аннотация

В связи с развитием и внедрением новых технологий, увеличением объема передаваемых данных все большую актуальность приобретает задача анализа сетевого трафика. Решение задачи анализа сетевого трафика можно разделить на три независимые подзадачи: захват, хранение и сам анализ. В данной статье описано программное обеспечение для захвата сетевого трафика. Оно включает предварительный анализ полученных данных и предоставляет возможность перехвата пакетов, выделения из них необходимых параметров, таких как время, длина пакета, флаги и прочее. В приложении представлена возможность извлечения из файлов этой информации и, собственно, анализ перехваченных пакетов либо в режиме реального времени, либо обработка данных, полученных из файла. После захвата сетевого трафика, благодаря применению разработанного программного обеспечения, проведена выборка времен поступления и отправки пакетов. Далее были проанализированы полученные данные, вычислены такие параметры, как первый, второй и третий начальные моменты, выборочная дисперсия, коэффициент вариации и асимметрия. Под анализом будет подразумеваться вычисление таких параметров, как первый, второй и третий начальные моменты, выборочная дисперсия, коэффициент вариации и асимметрия. Данные статистики позволяют судить о характере распределения интервалов. Например, коэффициент вариации показывает отличие трафика от пу-ассоновского потока и совместно с асимметрией позволяет судить о степени весомости хвостов распределений. В итоге по этим параметрам будет возможность получить значение задержки данного трафика. Также программа позволит сравнить теоретически полученные данные с данными, полученными практически. Моментные характеристики распределения временных интервалов между принимаемыми пакетами, рассчитаны с использованием формул математической статистики. Данные статистики позволяют судить о характере распределения интервалов. Распределение интервалов между пакетами трафика относится к распределениям с тяжелыми хвостами. Система массового обслуживания Н2/М/1 с учетом весомости хвоста входного распределения, даже при маленьких нагрузках дает задержку во много раз большую, чем классическая модель.

Ключевые слова: сетевой трафик, выборочная дисперсия, коэффициент вариации, асимметрия, начальные моменты, гиперэкспоненциальное распределение..

Литература

1. Statistical analysis of network traffic [Электронный ресурс]. Дата обращения: http://pi.314159.ru (01.17.2018).

2. Analysis of the network traffic in real time: a review of applications, approaches and solutions [Электронный ресурс]. Дата обращения: http://www.vanderboot.ru (01.17.2018).

3. Тарасов В.Н., Бахарева Н.Ф., Горелов Г.А., Малахов С.В. Анализ входящего трафика в три момента в распределении временных интервалов // Информационные технологии, 2014. С. 54-59.

4. Тарасов В.Н., Малахов С.В. Statistical data handling program of Wireshark analyzer and incoming traffic research // Proceedings of the Institute for System Programming, 2015. vol. 27(3). pp. 303-314.

5. Tarasov V.N., Bakhareva N.F., Kartashevskiy I.V., Malakhov S.V. Analysis of intervals between traffic packets on the SDN networks depending on the TCP window size // Problems of Infocommunications Science and Technology (PIC S&T), 2016. pp. 15-17.

Информация об авторах

Малахов Сергей Валерьевич, Поволжский государственный университет телекоммуникаций и информатики, доцент кафедры Программного обеспечения и управления в технических системах, доцент кафедры ПОУТС, к.т.н., г. Самара, Россия

Мезенцева Екатерина Михайловна, Поволжский государственный университет телекоммуникаций и информатики, доцент кафедры Программного обеспечения и управления в технических системах, доцент кафедры ПОУТС, к.т.н., г. Самара, Россия