RESEARCH OF WIDE AREA NETWORKS PERFORMANCE

L.I.Abrosimov, K.V. Sultanov, Maishan Sedaghati

(Moscow, Russia Moscow Power Engineering Institute (Technical University))

 

 

The computer network  is a technical system containing computers, connected by communication devices and channels of communication, which provides remote performance of information tasks of the users.

The productivity or performance of computer network as a system serving users, should be estimated  by the total number of information tasks, which are carried out by all devices included into the structure of computer network.

Computer networks carry out functions of transferring and data processing and carry out transformation of the requests to the messages and vice versa. Because the  requests and messages are functionally related to each other, in notation we shall use the term  transaction, which is defined by  two parameters: the volume of the data and the number of computing operations necessary for processing.

Because of different requirements and the use of  various degrees of decomposition at the description of processing transactions in computer networks, it is necessary   to use  several degrees of computer networks performance,  which first of all it is possible to arrange the concepts of computer networks performance as following:

- Computer network performance of elements,

- Computer network performance of nodes,

- Complex computer network performance,

- Working computer network performance,

- Peak computer network performance,

- Limiting of computer network performance for considered type of transactions at peak loading.

In the basis of the method  of calculation of computer network performance the method of contours is taken place [4], [5].

In the report the example of calculation of  WAN performance  with application of the developed methods is in detail considered.

The given WAN, submitted in figure 1, contains in its structure: servers 1H - 5H (in the model are substituted numbers 1-5), switching controllers 1 - 10, connected communication  channels, (numbers 6-27) and group of the subscribers 1G - 5G (numbers 28-31, 32-37, 38-45, 46 51, 52-57) .

           Figure 1- Given WAN structure

The subscribers 1G - 5G in a dialogue mode interacts with appropriate servers in the scheme:

1G-1H (contour q=1); 2G-5H (q=2); 3G-1H (q=3); 4G-1H (q=4); 5G-2H (q=5).

In the given WAN the initial variant of routes for all of contours was set by a principle " fixed (uniform)loading of communication channels  ".

   Figure 2. Diagram of contours in a given WAN

 

In the following example the values of service intensity are accepted:

6 1/с;    4 1/с ;    0,333 1/с.

According to the previous definitions, we can get  the following numerical values for the given  WAN performance.

·                    The WAN performance of MN nodes, in our example does not depend on contours and can be written  as following :

 1/с . The values MN are used as the initial data for the further calculations.

·         Complex  МС WAN performance is written in a line for each contour q and in our example for q=1 it is possible to write  as following :

 1/с .МС allows to calculate the minimal time of delivery transactions for each contour q.

·             Working WAN performance is determined at  and for our example:

{NWq} ={2,3,4,3,3}, NQ = 5, {} = {0,171; 0,179; 0,248; 0,225; 0,184} 1/c, =3,0991/c; {tWq} = {2,84; 2,595; 2,366; 2,916; 2,436} с

·              Peak WAN performance, is determined at  and for our example: {NPq}={4,6,8,6,6}, NQ = 5, {} = {0,136; 0,150; 0,190; 0,175; 0,139} 1/c, = 4,8461/c; {tPq} = {4,353; 3,667; 4,02; 4,634; 4,171} с

·            limiting of WAN performance at limiting time of reaction of system =7,2 с.

For our example :   {NLq} ={6,9,12,9,9}, NQ = 5, {} = {0,1056; 0,120; 0,140; 0,131; 0,1} 1/c,  = 5,469 1/c ;{tLq} = {6,456; 5,361; 6,511; 7,165; 7,028} с .

 

Figure 3. Shows the intensity of transactions being serviced in different contours from the respective number of  subscribers

Figure 4. Shows the total of the WAN performance depend on the number of subscribers (working WAN performance, peak performance and limiting performance).

 

   Subscribers in the given WAN (wide area networking ) use dialogue interactions with servers, therefore the following  functioning modes: working, peak and limiting depend on the number of active subscribers .

In the Working mode  {NWq} ={2,3,4,3,3}, means. 15 Subscribers

In the Peak mode  {NPq} ={4,6,8,6,6}, means. 30 Subscribers

In the limiting mode {NLq} ={6,9,12,9,9}, means. 45 Subscribers

The increase of the number of subscribers (from the point of view of subscribers) reduces the performance of the network (see figure 3) as it increases the number of transactions in the network and  the line  also increases ,in the same way an increase in number of subscribers (from the systems administrator’s point of view ) increases the sum performance of the network (see figure 4)

The conclusion

The offered method of calculation of  WAN performance allows:

- To give the developers and system administrators an opportunity to estimate quantitatively  functioning computer network performance;

- To have a quantitative estimation for various variants of organization of computer network during development and modernization computer network;

- To determine "bootlenec" places in computer network and to offer the proved solutions based on their elimination by providing of additional resources or realization of  reconfiguration  computer network .

 

References

[1] Методы автоматизированного проектирования систем телеобработки данных// Учеб. пособие для вузов/ В.А. Мясников, Ю.Н. Мельников, Л.И. Абросимов. - М.: Энергоатомиздат,- 1992. - 288 с.: ил.

[2] Абросимов Л.И. Основные положения теории производительности вычислительных сетей / Вестник МЭИ – 2001 - №4 – С 70 - 75

[3] Abrosimov L.I. Conception of creation of intelligent integrated management platform for heterogeneous computer networks on the basis of tensortransform modeling. International Conference DISTRIBUTED COMPUTER COMMUNICATION NETWORKS. Theory and Applications 4 - 8.11.1997 Tel-Aviv (Israel). Institute for Information Transmission Problems RAS, Tel-Aviv Universiti, M:. 1997 s.3 - 11.

[4] Абросимов Л.И. Анализ и проектирование вычислительных сетей: Учебное пособие - М.:, Изд-во МЭИ. 2000. - 52 с

[5] Lehr- und Uebungsbuch Telematik /Hrsg. Gerhard Kruger ; Dietrich Reschke. Autoren Leonid I. Abrosimov, Jorg Deutschmann, Werner Horn, Holger Reif, Dietrich Reschke, Jochen Schiller, Jochen Seitz. Muenchen; Wien:Fachbucherverl. Leipzig im Carl-Hanser-Verl., 2000.S 85-98(412 s)