Alexander Tishenko
Faculty of Computer Information Technologies and Automatics
Department of Automation and Telecommunication
Speciality: "Telecommunication systems and networks"
Theme of master's work: "Researching of Control Process in Multiservice Telecommunication Networks with Using Forecasting Models"
Scientific Supervisor: Ph.D. (in Engineering), Associate Professor Vladimir Bessarab

ABSTRACT

of the Master's Qualification Work
"Researching of Control Process in Multiservice Telecommunication Networks with Using Forecasting Models"

Content

1. A theme urgency
2. Communication of work with scientific programs, plans, themes
3. The purpose and problems of the research
3. 1 The work purpose
3. 2 Idea of work
3. 3 The main problems of the research
3. 4 The subject of the research
3. 5 The object of the research
3. 6 Methodology and methods of researches
4. Prospective scientific novelty of the received results
5. Practical value of the received results
6. Approbation of the work results
7. The review of developments and researches on a theme
7. 1 At the local level
7. 2 At the national level
7. 3At the global level
8. The description of the received and planned results of the work
Conclusions
Bibliography
Note

1 A theme urgency

2 Communication of work with scientific programs, plans, themes

3 The purpose and problems of the research

3.1 The work purpose

The improvement of provided services’ quality in Multiservice Telecommunication Networks.

3.2 Idea of work

The redistribution of network resources between different services with using of Forecasting Models.

3.3 The main problems of the research

To achieve the aim it is necessary to solve the following problems:

• To analyze the requirements of QoS services, which are provided in Multiservice Network, to a communication channel;
• To study the efficiency; of the algorithms of prediction of loading in Multiservice Telecommunication Networks;
• To analyze the control modes of Multiservice Network;
• To research the control processes with different models of forecasting of loading in Multiservice Telecommunication Networks;

3.4 The subject of the research

The subjects of the development are algorithms of prediction and the control methods of Multiservice Networks.

3.5 The object of the research

The objects of the development are processes of prognostic management of Multiservice Networks.

3.6 Methodology and methods of researches

The Qualification Work has the following methods of researches: analytical one and modeling.

4 Prospective scientific novelty of the received results

• For the first time the algorithm of differential prediction of loading at the communication channel is proposed;
• The Theory of forecasting has obtained the further development;
• The Theory of self-similar processes and fractals has obtained the further development.

5 Practical value of the received results

The using of the system of prognostic management of Multiservice Network will allow to improve the quality of providing of services to the clients at the cost of prioritization and traffic tunneling, as well as at the cost of backing of network resources. For a telecoms operator this system will give the opportunity to minimize the losses due to a fact that at the moments of overloading of the communication channel the traffic discards or the one with the lower index of correlation &ldquo monetary units per bit&rdquo or with other index, defined by a telecoms operator stands in a queue. Traffic prediction and tunneling gives opportunity to use the channel resource as much as it’s needed at the moment, and the residual resource can be used at the discretion of a telecoms operator.

6 Approbation of the work results

Within the bounds of the Qualification Work it is planned to speak on the 25-27th of May of 2010 at the 5th Theoretical and Practical Conference &ldquo Donbass 2020: The trends of development with the eyes of young scientists&rdquo with the report issue &ldquo The loading prediction in Multiservice Networks&rdquo

7 The review of developments and researches on a theme

7.1 At the local level

At University the issues of forecasting of loading in Telecommunication Network were studied by Soloviev M.S. At his publication &ldquo The traffic GSM prediction with a glance of fractality behaviors&rdquo were examined the loading forecast in the mobile communication network. The problems of providing QoS in Multiservice Network were studied by Fedoseeva O. S. At her Master’s Qualification Work &ldquo The research of peculiarities of providing of characteristics of the service quality of different traffic modes in NGN-multiservice Networks&rdquo were examined the mechanisms of providing QoS and were defined characteristics of service quality for different types of Network.

7.2 At the national level

At the national level within the bounds of the problem the following researches are being held:
The publication, Artemenko M. E., Kasymov R. R., The State University of ICT, at which were proposed the prediction of Multiservice Traffic with neural network using , meanwhile the prognostic meaning of the traffic were presented as the sum of the following indexes: trend, showing the growth of user base, the weighting of Network Applications and so forth; the periodic component; traffic fluctuation generated by the Network resources distribution, routers’ buffer overflow, temporal overloads, failures of Network Elements, etc.; casual traffic jitter at the short-term scale. Every component is predicted separately. To solve the problem of information decomposition about Network traffic into rectangular components it was proposed to employ discrete wavelet transform. The publication of Kudzinovskaya I. P. The Institute of Computer Technologies, The National Air University. Were analyzed the methods of providing the service quality in the Networks. Also were examined the influence of self-similarity behavior of traffic at the indexes of QoS. Also were justified the necessity of logical design of traffic formation and the overload prevention with a glance of the traffic self-similarity [2].

7.3 At the global level

The dissertation of Repin D. S. on the theme &ldquo The analysis and modeling&rdquo solvs such problems: were worked out the complex methodology of the experimental research of the traffic in the Communication Networks; were worked out the method of simulation of the super intensity line of the non-stationary traffic. [3]. The publication of Aliev R.T. &ldquo Methods of traffic control in Multiservice Networks&rdquo. Were proposed the mechanism of traffic priority and its further treatment at the communications centers. The optimization criterion is a minimization of the burst buffer facility delay. The dissertation of Platov V. V. on the theme The dissertation of Repin D. S. on the theme &ldquo The analysis and modeling&rdquo solvs such problems: were worked out the complex methodology of the experimental research of the traffic in the Communication Networks; were worked out the method of simulation of the super intensity line of the non-stationary traffic. [3]. The publication of Aliev R.T. &ldquo Methods of traffic control in Multiservice Networks&rdquo. Were proposed the mechanism of traffic priority and its further treatment at the communications centers. The optimization criterion is a minimization of the burst buffer facility delay. The dissertation of Platov V. V. on the theme &ldquo The special Mathematical and Software of the control processes of intensity of data transmission&rdquo, in which were researched the algorithm of reasonable choice of the technology of forecast of self-similar processes, providing the choice of the most efficient with a view to the accuracy and method’s computational complexity of forecasting of the self-similar processes in Comunication Systems. Also were proposed the control algorithm of intensity of data flow, providing the optimization of data transmission at Computer Networks with the self-similar traffic &ldquo The special Mathematical and Software of the control processes of intensity of data transmission&rdquo, in which were researched the algorithm of reasonable choice of the technology of forecast of self-similar processes, providing the choice of the most efficient with a view to the accuracy and method’s computational complexity of forecasting of the self-similar processes in Comunication Systems. Also were proposed the control algorithm of intensity of data flow, providing the optimization of data transmission at Computer Networks with the self-similar traffic

8 The description of the received and planned results of the work

The Multiservice Network is a Network, where a client can receive some different services in the one user line (services combination):

• Internet – access to the Worldwide Network, Web-pages, FTP, e-mail, etc.;
• VPN – virtual private network that allows to construct networks or secure channels via Internet (or rent channels of third parties);
• VoIP – IP-telephony, telephone communications, based on the packet switching;
• IPTV – the transmission of video picture IP networking .

Each of these services sets up its claims to communication channel for the sake of full value functioning:

Table 8.1 – Requirements of QoS of different Services

Where,
t_c – time of setting up, sec;
p^(rj) – connection release probability;
dT – delay, msec;;
Dj – jitter, msec;
B – channel bandwidth.
Since the physical channel is the one, and the service requirements are various, then the channel resource distribution between services is the important function for the QoS services requirements providing [2].

Also it is worth saying that the traffic generated by these applications has the property of self-similarity. The random process X(t) is considered as self-similar with Harst characteristics H > 0, if statistics of the process X(t) don’t change while scaling in-phase on a^-h and with timing on a for all а > 0:

Where,
H – is a Harst coefficient or characteristics that represents self-similarity measure.

For self-similar processes the following properties are typical:

• Dispersion of aggregated plot with a aggregation coefficient m is:

  Var[X(m)]=Var[X]/mb

  (8.2)

  
  

  b=2(1-H)

  (8.3)

  
  

  Where,
  Var[x] – is a process dispersion;
  Var[x^(m)] – – an aggregative process dispersion ;
  m – is a coefficient og aggregation;

• autocorrelation function of the aggregative self-similar process doesn’t tend to zero with m to infinity.

Self-similar processes can have a property of long-time dependence, that means the dependence demonstration between the events in a long enough term [3]. Not the least of the properties of the traffic in Multiservice Networks is the property of periodicity, that is caused by a factor that the loading level into the Network depends on daily users activity. At night the Network loading is commonly lower, and at the daytime it may be singled out or several activity peaks.

This property is important for making forecasts of Network loading. For this it may be distinguished out several periods: daily, weekly. Due to it it may be estimated the average level of loadind at the definite period of time.

To predict the the loading may be used the following standart algorithms of forecasting:

• moving-average process;
• exponential smoothing process;
• double exponential smoothing process;
• triple exponential smoothing process;
• autoregressive methods (AR, ARMA, ARIMA);
• neural network prediction [4];

These algorithms while using per se has a big accuracy. For more exact result it may be used the following algorithm:

The prognostic meaning of loading the next moment of time is:

(8.4)

Where,
В – is a prognostical meaning of the traffic at the next period of time;
NextAvg – is a expectation value at the next moment of time;
NextDisp – is a dispersion at the next moment of time;

According to a property of traffic periodicity, let’s form the time series from the definite number of counts which are took from the main Network statistics incrementally. According to a formed row we can predict the expectation value at the next moment of time with one of the standard forecast processes, for instance, with the exponential smoothing process.

The dispersion for the next period of time may be calculated on the assumption of the properties of self-similarity of traffic in Multiservice Networks. For the aggregative traffic realization included at the watch window it is necessary to calculate tha Harnst coefficient. Then, using the formulae 8.2 and 8.3 we may calculate the process dispersion at the next period of time:

Where,
Disp – is a dispersion of the aggregative process.

To redistribute the channel resources the traffic in the Network must be prior (every package must have the mark to which type of traffic it belongs). Due to it it is possible to hold the process separately for each type of traffic in the channel. If the increasing of traffic level is expecting with a definite priority, then the Network recourses may be distributed what will allow to prevent the losses of the definite type of traffic.

On of the ways of resources redistribution is the traffic tunneling (at each tunnel it is putdown the traffic with the same priority of the mechanisms of the technology MPLS) and the backing of the Network resources by means of protocols RSVP. If the system of forecasting supposes the growth of one of the types of traffic, the redistribution of Network resource is made (in our case the channel capacity) according to the following algorithm:

1. If there are non reserved resources in the channel, they may be reserved in accordance with this type of traffic;
2. If there are no free resources in the channel, but there is an alternative route of delivery of the packets with worse metrics and there is a free resources in it, we can construct the tunnel through it and through this tunnel begins the transportation of a part or of the whole traffic with the least priority in the channel. As for the tunnel for this traffic at the main channel decreases or completely removes.
3. If there are no free resources in the channel as well as there are no alternative routes, the tunnels with the traffic of less priority are going decreased at the required quantity.
Due to such redistribution of resources we satisfy one of the service requirements to a quality of service, that is communication channel bandwidth.

Figure 8.2 - An example of resource allocation between different types of traffic in the channel

(Animation: volume - 13 КB; size - 372х176; delay between the last and first shots - 500 ms; number of repetition cycles - infinite)

Conclusions

Due to using of control traffic prognostic system in Multiservice Telecommunications Network there appear an opportunity to redistribute resources in the Network, which allows to satisfy the requirements of services at the service quality, thereby providing the call-off quality of providing services to the clients in multiservice Network.

Bibliography:

1. Global Bandwidth Research Service [Electronic resource] - Access mode to article: http://www.telegeography.com/product-info/gb/index.php
2. Величко В.В. Телекоммуникационные системы и сети: Учебное пособие. В 3 томах. Том 3. - Мультисервисные сети. / Е.А. Субботин, В.П. Шувалов, А.Ф. Ярославец. - М.: Горячаяя линия - телеком, 2005. - 592 с.
3. Столлингс В. Современные компьютерные сети. - [2-е изд]. - СПб.: Питер, 2003. - 783 с.
4. Дуброва Т.А. Статистические методы прогнозирования: Учеб. пособие для вузов. - М.: ЮНИТИ-ДАНА, 2003. - 206 с.
5. Громов Ю.Ю. Фрактальный анализ и процессы в компьютерных сетях : учеб. пособие. - [2-е изд]. Н.А. Земской, О.Г. Иванова, А.В. Лагутин, В.М. Тютюнник. - Тамбов: Тамб. гос. техн. ун-та, 2007. - 108 с.
6. Лившиц Б.С. Теория телетрафика. Учебник для вузов. - [2-е изд]. / А.П. Пшеничников, А.Д. Харкевич. - М.: Связь, 1979. - 224с.
7. Кошманский В.И. Нейронные сети и их применение в системах управления и связи. / Д.А. Смирнов. - М.: Горячая линия - Телеком, 2003. - 94 с.
8. Крылов В.В. Теория телетрафика и её приложения. / С.С. Самохвалова. - СПб.: БХВ-Петербург, 2005. - 288 с.
9. Вегешна Ш. Качество обслуживания в сетях IP. - М.: Издательский дом "Вильямс", 2003. - 368 с.
10. Филимонов А.Ю. Построение мультисервисных сетей Ethernet. - СПб.: БХВ-Петербург, 2007. - 592 с.

Notes

While making this abstract the Master’s Qualification Work hasn’t been completed yet. The date of the final completion is the 20th of December, 2010. The full work content as well as all theme materials may be received from the author or his scientific adviser after the mentioned date.

DonNTU Master's portal