20.11.2012 Views

Contents Telektronikk - Telenor

Contents Telektronikk - Telenor

Contents Telektronikk - Telenor

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

26<br />

n *<br />

ooo---o<br />

A * M, V<br />

stream is not identical to a single overflow<br />

stream with exactly the same mean<br />

and variance. However, for practical<br />

application the overflow given by the<br />

pair (M,V), as calculated above, is assumed<br />

to be equivalent to a single overflow<br />

stream with the same (M,V) - pair. The<br />

principle is shown in Figure 28, where M<br />

and V are calculated sum values and the<br />

pair (A*,n*) represents the equivalent<br />

primary group to be determined. Riordan’s<br />

variance formula gives an explicit<br />

expression A* = f(n*) or n* = g(A*). In<br />

principle A* and n* are found by insertion<br />

in Erlang’s formula. This formula,<br />

however, is only implicit, and the solution<br />

can only be found by an approximation<br />

formula, by tables or by numerical<br />

calculation.<br />

When A* and n* are determined, the<br />

overall lost traffic is given by m:<br />

m = A* ⋅ E (n* + k, A*)<br />

and the loss probability<br />

g�<br />

B = m/A = m/<br />

Note that the loss B is the ratio between<br />

the calculated lost traffic and the total<br />

offered real traffic, m/A, and not related<br />

to the fictitious traffic A* by m/A*.<br />

The method is called the ERT (Equivalent<br />

Random Traffic) method and often<br />

referred to as Wilkinson’s equivalence<br />

method [10]. It can be applied in a stepwise<br />

manner for progressive gradings.<br />

Another, and at present more interesting<br />

area, is that of networks with alternative<br />

routing, where primary circuit groups<br />

k<br />

ooo---o m<br />

Figure 28 A fictitious equivalent group for generation of a given overflow M,V<br />

A 1primary<br />

A 2primary<br />

n 1<br />

ooo---o<br />

⇓<br />

A 1overflow<br />

Reserve part of n 2 forA 2primary<br />

Figure 29 Trunk reservation for primary traffic in order to balance loss<br />

i=1<br />

Ai<br />

n 2<br />

oooooo---<br />

may have overflow to a common secondary<br />

group.<br />

14.5.2 The IPP (Interrupted<br />

Poisson Process) method<br />

Another method for overflow analysis<br />

was introduced by Kuczura [11], and is<br />

based on the distinction between the two<br />

states S1 = [Primary group is busy] and<br />

S2 = [Primary group not busy]. With the<br />

primary input being a Poisson process,<br />

the secondary group will see a switching<br />

between Poisson input and no input. The<br />

method is thus named IPP = Interrupted<br />

Poisson Process. The approximation is<br />

the assumption that the dwelling times in<br />

the two states are exponential, which we<br />

know is not true. The process has essentially<br />

three parameters, the Poisson<br />

parameter λ, and the S1 and S2 exponential<br />

parameters α and β. A three-moment<br />

match is based on the mentioned three<br />

parameters. An alternative method by<br />

Wallstrøm/Reneby is based on five<br />

moments in the sense that it uses the first<br />

moment and the ratios of 2nd/3rd and<br />

7th/8th binomial moments. [12].<br />

The IPP method is widely used and has<br />

been generalised to the MMPP (Markov<br />

Modulated Poisson Process) method,<br />

where the switching occurs between two<br />

non-zero arrival rates.<br />

The methods presented do not assume that<br />

all primary groups are equal or near equal.<br />

The independence assumption assures the<br />

correctness of direct summation of means<br />

and variances. Thus, calculation of the<br />

overall loss is well founded. In symmetrical<br />

cases the overall loss B as calculated<br />

above will apply to each of the single pri-<br />

⇓<br />

mary groups. However, with substantial<br />

non-symmetry the single groups may<br />

experience quite different loss ratios. This<br />

has been studied by several authors, but<br />

will not be treated in this paper. Reference<br />

is made to [13], this issue.<br />

A particular dimensioning problem of<br />

overflow arises in network cases when<br />

overflow from a direct route to a final<br />

route mixes with a primary traffic on the<br />

final, Figure 29. The A2 traffic may experience<br />

a much higher blocking than A1 .<br />

The problem can be solved by splitting<br />

n2 and reserving part of it to A2 only, to<br />

obtain a more symmetric grading.<br />

14.6 Access limitations by link<br />

systems<br />

Link systems were developed on the<br />

basis of crossbar switches and carried on<br />

for crosspoint matrices in early electronic<br />

switching systems. Like many grading<br />

methods, link systems had their motivation<br />

in the limitations of practical constructions,<br />

modified by the need for<br />

improved utilisation of quite expensive<br />

switching units.<br />

This can be explained as follows: A traditional<br />

selector has one single inlet and,<br />

say, 100 to 500 outlets. When used as a<br />

group selector, 10 to 25 directions with<br />

10 to 20 circuits per direction would be<br />

feasible. All this capacity would still only<br />

permit one connection at a time, and a<br />

large number of selectors with common<br />

access to the outlets (in what is termed a<br />

multiple) would be necessary to carry the<br />

traffic. A crossbar switch could be organised<br />

in a similar way with one inlet and,<br />

say, 100 to 200 outlets, and with the<br />

same limited utilisation. However, the<br />

construction of the crossbar switch (and<br />

likewise a crosspoint matrix), permits a<br />

utilisation whereby there are, say, 10<br />

inlets and 10 – 20 outlets, with up to 10<br />

simultaneous connections through the<br />

switch. This tremendous advantage carries<br />

with it the drawback of too few outlets<br />

for an efficient grouping. The solution<br />

to this is a cascading of switches in 2<br />

– 3 stages. That, of course, brings down<br />

the high efficiency in two ways: an increase<br />

in the number of switches, that<br />

partly “eats” up some of the saving, and<br />

introduction of internal blocking, with a<br />

corresponding efficiency reduction. Still,<br />

there is a considerable net gain.<br />

The selection method through two or more<br />

stages is called conditional selection, as<br />

the single link seizure is conditioned on<br />

the whole chain of links being available,<br />

and so are seized simultaneously.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!