WWW.DISSERS.RU


...
    !

Pages:     | 1 |   ...   | 50 | 51 || 53 | 54 |

Fo, Tb, S1, S2, S3, S1Z, S2Z, R1, R2, R3 (3.2) Aim: To determine the number of switching lines Ns; and the following unknown parameters:

Yab, Fa, dem.Fa, rep.Fa, ofr.Fs, Ts, ofr.Ys (3.3) Condition:

Pbl ( Pbr, Pbs) adm.Pbl (3.4) 4.2. Solvability of the NDT:

The traffic intensity Yab characterizes the macrostate of the system. In Poryazov, Saranova (2005) is shown that F0(S1 - S3Pbs)-(F0(S1 - S3Pbs)+ F0S2(1- Pbs))Pbr + F0S2(1- Pbs)PbrYab = (3.5) F0(S1 - S3Pbs)-(F0M(S1 - S3Pbs)+ F0S2(1- Pbs)-1+ R1 - R3Pbs)Pbr + (1- Pbs)(F0MS2 + R2)PbrTheorem 1: If Pbr 0 and Fo 0, then analytical presentation (3.8) of Yab in the NDT exist.

Proof: Considering the system equations (1.1) - (1.10) when Pbr 0 and Fo 0 from (1.5) and (1.3) follows Fo dem.Fa = [ Pbr -1+ (M Pbr +1)Yab] (3.6) Pbr From (1.2) and (1.4) follows dem.Fa = Fa {1- R1 + R2 Pbr + (R3 - R2 Pbr) Pbs} (3.7) Then (3.6), (3.7) and (1.2) gives F0(1- Pbr){S1 - S2Pbr -(S3 - S2Pbr)Pbs} Yab = (3.8) F0(1+ MPbr){S1 - S2Pbr -(S3 - S2Pbr)Pbs}- Pbr{1- R1 + R2Pbr + (R3 - R2Pbr)Pbs} (3.8) is new simplified expression of the (3.5).

If Fo = 0, then obviously Fa = 0, dem.Fa =0 and rep.Fa = 0.

Therefore, when Pbr 0 in NDT, on the base of administrative determined values of parameters Pbs, Pbr, M and the known parameters (3.2), traffic intensity Yab is derivable. The other system parameters in the NDT are depending on the system state (respectively on Yab).

Fourth International Conference I.TECH 2006 We will prove that the values of unknown parameters (3.3) in the NDT can be derived (evaluated) through Yab and known parameters (3.2) in correspondence of determined conditions.

Theorem 2: If S1 - S2Pbr Pbr 0 and Pbs (3.9) S3 - S2Pbr in the NDT, then for each unknown parameter of (3.3), an analytical expression for its evaluation exists.

Proof: Using the system (1.1) (1.10) and from (1.1) and (3.5) by (S1 S2 Pbr) - (S3 S2 Pbr) Pbs 0, follows Yab Fa = (3.10) S1- S2 Pbr -(S3- S2 Pbr) Pbs For dem.Fa from (1.3) and (1.5) is received (3.6).

It is resulted from (1.4) and (3.9):

Yab{R1- R2 Pbr -(R3- R2 Pbr) Pbs} rep.Fa =.

(3.11) S1- S2 Pbr - (S3- S2 Pbr) Pbs From (1.6) and (3.9) follows:

Yab (1- Pad )(1- Pid ) ofr.Fs = (3.12) S1- S2 Pbr - (S3- S2 Pbr) Pbs The parameter Ts can be calculated from (1.7), and from (1.3) and (1.5) follows:

(1- Pad )(1- Pid )(S1z - S2z Pbr)Yab ofr.Ys = (3.13) S1- S2 Pbr - (S3- S2 Pbr) Pbs Therefore, the values of the unknown parameters (3.3) in the NDT can be expressed and calculated by the conditions of the Theorem 1 and Theorem 2.

For the network dimensioning, when the level of QoS is determined administratively in advance (for example blocking probability Pbs), ErlangsB - formula may be used:

Pbs = Erl_b(Ns,ofr.Ys) (3.14) (ofr.Ys)Ns Ns! Erl _ b(Ns,ofr.Ys) =.

(3.15) j Ns (ofr.Ys) j! j =The number of switching lines Ns and the values of ofr.Ys are calculated by the conditions of the Theorem 1 and Theorem 2.

Remark 1-2: ofr.Ys, being evaluated on the base of the Theorem 1 and Theorem 2 for adm.Pbs and adm.Pbr, is resulted in a fixed value. Then Pbs = Pbs (Ns,ofr.Ys) is a function of Ns only and Pbs = Pbs (Ns).

Theorem 3: The function Pbs = Pbs (Ns,ofr.Ys), defined through (3.15) in the NDT is strictly monotone decreasing according to Ns 1, when ofr.Ys >0 is a fixed value.

Proof: It can be proved that Pbs (Ns+1, ofr.Ys) < Pbs (Ns, ofr.Ys). Obviously (see (3.15) Pbs (0,ofr.Ys) = 1.

Using the recursion ErlangsB - formula [Iversen 2004]:

Networks and Telecommunications ofr.Ys Pbs (Ns -1,ofr.Ys ) Pbs (Ns,ofr.Ys) =.

(3.16) Ns + Pbs (Ns -1,ofr.Ys ) But Erl_b(Ns,ofr.Ys) > 0 when ofr.Ys >0 and Ns 1, Ys Erl_b(Ns,ofr.Ys)+Ns+1> 0 and Pbs(Ns +1,ofr.Ys) - Pbs(Ns,ofr.Ys) = Erl _ b(Ns +1,ofr.Ys) - Erl _ b(Ns,ofr.Ys) = ofr.Ys[1- Erl _ b(Ns,ofr.Ys)] - (Ns +1) = Erl _ b(Ns,ofr.Ys) = ofr.Ys Erl _ b(Ns,ofr.Ys) + (Ns +1) crr.Ys - (Ns +1) Erl _ b(Ns,ofr.Ys).

ofr.Ys Erl _ b(Ns,ofr.Ys) + (Ns +1) Because crr.Ys Ns follows crr.Ys (Ns+1) < 0.

Therefore, Pbs ( Ns+1, ofr.Ys) Pbs ( Ns, ofr.Ys) < 0 and the function Pbs = Pbs (Ns,ofr.Ys), defined through (3.14) is strictly monotone decreasing, when ofr.Ys > 0 is fixed value.

5. Analytical Solution Based on the Assumption A-8 we are working with mean values of the parameters. Various techniques for analyzing complex teletraffic systems require a formulation of the Erlang function that is continuous in the parameter Ns. This is done via the integral representation [Berezner 1998].

Theorem 4: There is only one solution in the NDT through the equation Erl_b (Ns,ofr.Ys) = adm.Pbs, (5.1) according to the number of switching lines Ns.

Adm.Pbs (0; 1] is in advance administratively determined value of blocking probability, providing of QoS.

Proof: Existence: It was proved, that the function Pbs = Pbs (Ns,ofr.Ys), defined through (3.14) in the NDT, is strictly monotonic decreasing, when ofr.Ys >0 is fixed value. The number 1 is absolute maximum and 0 is absolute minimum of the function. There is only one solution of the equation (3.15) for adm. Pbs (0; 1], relying of the Intermediate Value Theorem (Dirschmidt, H. Yorg, 1992).

Uniqueness: Admitting that there are two different solutions Ns1 Ns2 of the equation (11.19) for adm. Pbs (0;

1], therefore they are simultaneously fulfilled Erl_b(Ns1, ofr.Ys) = adm.Pbs and Erl_b(Ns2, ofr.Ys) = adm.Pbs, is contradicting to Theorem 3.

It is proved that only one solution of Ns exist, fulfilling the equation (3.15) and corresponding to the determined administratively in advance value of the blocking probability adm.Pbs (0; 1].

6. Algorithm for Calculating the Values of the Parameters in the NDT:

1. From SLA and ITU-Recommendation are specified and determined administratively blocking probability adm.Pbl, respectively adm. Pbs (0; 1] and adm.Pbr [0; 1]:

adm.Pbl (adm.Pbr, adm.Pbs) :

(6.1) adm.Pbr adm.Pbs = adm.Pbl : adm. Pbr [0,1], adm. Pbs (0,1] 2. The unknown parameters (3.3) in the NDT are evaluated on the base of Theorem 1 and Theorem 2, known parameters (3.2), especially adm.ofr.Ys (adm.Pbr, adm.Pbs).

3. On the basis of each calculated value adm.ofr.Ys, we evaluate ~ ~ ~ Ns R+ :{(adm.ofr.Ys, Ns) : Pbs (adm.ofr.Ys, Ns) = adm.Pbs} (6.2) Fourth International Conference I.TECH 2006 ~ 4. If adm.Ns = sup Ns, then Ns =[adm.Ns]+1: Pbl adm.Pbl.

(6.3) 5. For finding of the number of internal switching lines Ns, a computer program is created on the base of the recursion ErlangsB formula (6.15) [Iversen 2004]. From numerical point of view, the following linear form is the most stable:

Ns I (Ns,ofr.Ys) = 1+ I (Ns -1,ofr.Ys), I (0,ofr.Ys ) = 1, (6.4) ofr.Ys where I(Ns,ofr.Ys) = 1/ Pbs(Ns,ofr.Ys). This recursion formula is exact, and for large values of (Ns, ofr.Ys) there are no round of errors.

6. The received results for numerical inversion of the Erlangs formula (for finding the number of switching lines Ns) were confirmed with results of others commercial computer programs.

Therefore, it is proved that if Pbr 0 and Pbs (S1 S2 Pbr) / (S3 S2 Pbr), then the NDT is solvable and there is proposed algorithm for its solution.

When Pbr = 0 the network loading is rather low and it is not of great practical interest, but in this case a mathematical research is made also.

7. Numerical Results Among the easy computable QoS - parameters in the system (resulted from QoS- strategy of the network operators) is blocking probability Pbl in pie-form model [Poryazov 2000]. The sum of the loss probabilities due to abandoned and interrupted dialing, blocked and interrupted switching, not available intent terminal, blocked and abandoned ringing and abandoned conversation in pie- form model is 1.

For finding of the main teletraffic characteristics in proposed conceptual and its corresponding analytical model, the so called normal-form model (see Fig. 1) is used for presentation of blocking switching probability (Pbs) and probability of finding B-terminal busy (Pbr).

0.Pbl 0.0.0.0.0.0.0.0.0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Ox: Pbr Fig. 2. General blocking probability Pbl in pie-form model is presented as function of probability of finding B-terminal busy Pbr and probability of blocking switching Pbs in normal-form Pbr and Pbs in 3D and contour - map presentation.

Oy: Pbs Oz: Pbl P b l r O b y P :

:

P x b O s Networks and Telecommunications Based on the conceptual and its corresponding analytical model (1.1)- (2.9), defined general blocking probability Pbl is presented in pie-form model in (2.10) as function of the Pbr and Pbs (which are presented in normal- form model in the same equation).

On the Fig. 2 blocking probability Pbl is shown in pie- form model, depending on probability of finding B-terminal busy Pbr (Ox axis) and probability of blocking switching Pbs (Oy axis) in normal- form model. Pbl increases when Pbr and Pbs increase. Therefore, when adm.Pbl is predetermined as level of QoS administratively then adm.Pbr and adm.Pbs can be determined (evaluated) correspondingly.

Ns/ Nab 0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Ox: Pbr Fig. 3. The number of equivalent internal switching lines Ns (as percentage of number of active terminals Nab, where Nab= 7000 terminals) and general blocking probability Pbl are shown as functions of Pbr (Ox axis) and Pbs (Oy axis) in normal-form model, as well.

Conclusions of the numerical experiments:

According Pbr, Pbs and Pbl:

If Pbr [0;1] and Pbs [210-9;0.999917] then 1. Pbl [0;0.900896].

Ns 2. 0.000143 0.387857, Ns [1;2715], when Nab = 7000 ;

Nab ofr.Ys 3. 0.77 10-5 1.728311, ofr.Ys [ 0.473782;12098.18], when Nab = 7000.

Nab Ofr.Ys may exseed Nab by 73% approximately. This is unproductiveness occupying of resources.

4. Absolute maximum for ofr.Ys:

Maximum ofr.Ys = 12098.18 and this value is about 4.9 times greater than switching system capacity Ns = 2715.

Absolute maximum for Ns:

Ns = 2715 when Nab=7000 terminals, Ns = 38.79% of Nab. This is possible if Pbl = 0.900882 90% (maximum theoretical value of Pbl ), Pbr= 0.876623 87.7% and Pbs = 9.2810-9, Yab= 6136.487 Erl 87.66 % of Nab and ofr.Ys = 2452.021 Erl = crr.Ys 35.0289 % of Nab.

Oy: Pbs Oz: Ns/ Nab Pbl r b P x:

O O y :

P b s Fourth International Conference I.TECH 2006 0.0.0.0.0.0.Ns/ Nab 0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Ox: Pbr Fig. 4. The offered traffic ofr.Ys, the number of internal switching lines Ns and general blocking probability Pbl are presented as function of probability of finding B-terminal busy Pbr and probability of blocking switching Pbs in normal form model in 3D and contour- map presentation.

8. Conclusions 1. Detailed normalized conceptual model, of an overall (virtual) circuit switching telecommunication system (like PSTN and GSM) is used. The model is relatively close to the real-life communication systems with homogeneous terminals.

2. General blocking probability Pbl as GoS parameter and adm.Pbl as QoS parameter in pie - form model are formulated. The offered traffic ofr.Ys, the number of internal switching lines Ns and general blocking probability Pbl are derived as functions of probability of finding B-terminal busy Pbr and probability of blocking switching Pbs in normal form model.

3. The network dimensioning task (NDT) is formulated on the base of preassigned values of QoS parameter adm.Pbl and its corresponding GoS - parameters - adm.Pbr and adm.Pbs; The NDT is formulated on condition that Pbl adm.Pbl.

4. The conditions for existence and uniqueness of a solution of the NDT are researched and an analytical solution of the NDT is found;

5. An algorithm and a computer program for a calculation the values of the offered (ofr.Ys), carried (crr.Ys) traffic and the number of equivalent switching lines Ns, are proposed. The results of numerical solution are derived and graphically shown;

6. The received results, in NDT, make the network dimensioning, based on QoS requirements easily;

7. The described approach is applicable directly for every (virtual) circuit switching telecommunication system (like GSM and PSTN) and may help considerably for ISDN, BISDN and most of core and access networks dimensioning. For packet switching systems, like Internet, proposed approach may be used as a comparison basis especially when they work in circuit switching mode.

Oy: Pbs Oz: ofr.Ys / Nab of r.

Ys / b a Nab N / s Y.

r f o Pb l r O b P y :

:

x P b O s Networks and Telecommunications References Berezner S.A., 1998, - On the inverse of Erlangs Function- J. Appl. Prob.35, 246- Dirschmidt, Hans Yorg, 1992 - Mathematische Grundlagen der Elektrotechnik-Braunschweig, Wiesbaden, pp.Engset, T., 1918. The Probability Calculation to Determine the Number of Switches in Automatic Telephone Exchanges.

English translation by Mr. Eliot Jensen, Telektronikk, juni 1991, pp 1-5, ISSN 0085-7130. (Thore Olaus Engset (18651943). "Die Wahrscheinlichkeitsrechnung zur Bestimmung der Whlerzahl in automatischen Fernsprechmtern", Elektrotechnische zeitschrift, 1918, Heft 31.) ITU E.501. ITU-T Recommendation E.501: Estimation of traffic offered in the network. (Previously CCITT - Recommendation, revised 26. May 1997) ITU E.600, ITU-T Recommendation E.600: Terms and Definitions of Traffic Engineering (Melbourne, 1988; revised at Helsinki, 1993).

ITU E.800. ITU-T Recommendation E.800: Terms and Definitions related to Quality of Service and Network Performance, including Dependability. (Helsinki, March 1-12, 1993, revised August 12, 1994).

Iversen V. B., 2004. Teletraffic Engineering and Network Planning, Technical University of Denmark, pp.Little J. D. C., 1961. A Proof of the Queueing Formula L=W. Operations Research, 9, 1961, 383-387.

Poryazov S. A, Saranova E. T., 2002. On the Minimal Traffic Measurements for Determining the Number of Used Terminals in Telecommunication Systems with Channel Switching. In: Modeling And Simulation Environment for Satellite and Terrestrial Communication Networks - Proceedings of the European COST Telecommunications Symposium, Kluwer Academic Publishers, 2002, pp. 135-144;

Pages:     | 1 |   ...   | 50 | 51 || 53 | 54 |



2011 www.dissers.ru -

, .
, , , , 1-2 .