20.11.2012 Views

Contents Telektronikk - Telenor

Contents Telektronikk - Telenor

Contents Telektronikk - Telenor

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

t1 t2 t3 t4 t5 t6<br />

T1 T2 T3 T4 T5 T6 T7<br />

equal size as the one around 10. One<br />

observation is that the residential subscribers<br />

cause a better utilisation of the<br />

system than do the business subscribers,<br />

due to the greater traffic concentration of<br />

the latter.<br />

An example of data traffic observations<br />

[3] is shown in Figure 10, a), b), and c).<br />

The observations are given in the form of<br />

bits per second from a workstation over a<br />

three hour period. Three degrees of resolution<br />

are given, 1 second, 10 seconds<br />

and 1 minute, clearly indicating the loss<br />

of detail by integration over longer time.<br />

Indicative is the reduction of peak values<br />

from 750 via 350 to 160 kb/s for the<br />

same set of data. (Note different ordinate<br />

scales.)<br />

8 Traffic modelling<br />

τ1 τ2 τ3 τ4 τ6 τ5<br />

s1 s2 s3 s5<br />

Figure 11 The double process of arrivals and departures<br />

Up till now we have focused on the traffic<br />

load on a set of servers and on different<br />

types of variation. Apart from an<br />

indication that the traffic is generated by<br />

calls from traffic sources we have not<br />

studied how traffic is created.<br />

In Figure 3 the call arrivals create all the<br />

+1 steps. In the long run there are equally<br />

many -1 steps, and these all stem from<br />

departures from the system. Earlier or<br />

later any arrival must result in a departure.<br />

This double process may be depicted<br />

as in Figure 11. If we start by observing<br />

only arrivals, we have the points T1 to T7 with the intervals t1 to t6 .<br />

Arrivals and departures are always connected<br />

in pairs. This implies that the longterm<br />

averages of arrival rate λ and departure<br />

rate γ must be equal. That again<br />

secures a limitation of the traffic load as<br />

long as λ and the holding times (service<br />

times si ) are limited. It is seen from the<br />

figure that the traffic measured during the<br />

period T = T7 – T1 is a little less than 1<br />

Erlang. This is seen by adding the holding<br />

times and dividing the sum by T: A = (s1<br />

s4 s6<br />

+ s2 + s3 + s4 + s5 + s6)/T. By reducing<br />

the intervals T i – T i–1 , thereby increasing<br />

λ, the traffic increases. The same effect is<br />

obtained by increasing s i .<br />

9 Little’s formula, the<br />

“Ohm’s law” of teletraffic<br />

It is well known from electrophysics that<br />

the three quantities voltage v, current i<br />

and resistance r are connected by the formula<br />

v = i ⋅ r, and that in an electric network<br />

this formula may be applied to the<br />

whole network or any part of it. This is<br />

Ohm’s law. In a similar way the three<br />

quantities traffic A, arrival rate λ and<br />

service (holding) time s are connected by<br />

the formula<br />

A = λ⋅s (3)<br />

This is Little’s formula, and like Ohm’s<br />

law in an electric network, Little’s formula<br />

applies to the whole or any part of a<br />

traffic network. A difference is that for<br />

Ohm’s law constant values are assumed,<br />

whereas Little’s formula applies to mean<br />

values, including the constant case (given<br />

co-ordination of arrivals and departures,<br />

which is a bit artificial). The only condition<br />

is a stationary process, which will<br />

be discussed later. In simple terms stationarity<br />

means that the statistical properties<br />

of the involved processes remain<br />

unchanged over time. There are no other<br />

conditions laid on the statistical distributions<br />

of the variables.<br />

With reference to the traffic models of<br />

Figures 1 and 2 an alternative model is<br />

shown in Figure 12.<br />

The following indexes can be used:<br />

o ⇒ offered, relating to sources<br />

l ⇒ lost<br />

q ⇒ queue<br />

c ⇒ carried, relating to servers.<br />

One could also use corresponding indexing<br />

on holding times, however we use,<br />

according to common practice, the following:<br />

s ⇒ service time<br />

w ⇒ waiting time (in queue).<br />

Very often h is used for holding time. It<br />

can be indexed to show which system<br />

part it is related to. For instance hl may<br />

be used to indicate that even a lost call<br />

occupies a source for a non-zero time.<br />

According to Little’s formula we obtain:<br />

Traffic load on queue:<br />

Aq = λq ⋅ w = λc ⋅ w<br />

Traffic load on server group:<br />

Ac = λc ⋅ s<br />

Traffic load on source group (encompassing<br />

the whole system):<br />

Ao = (λo – λl ) ⋅ (w + s) + λl ⋅ hl = λc ⋅ (w + s) + λl ⋅ hl We see that, if the lost traffic holding<br />

time hl is zero, the traffic load on the<br />

sources, Ao , is actually equal to the sum<br />

of the queue traffic and the server traffic.<br />

In this case – but not in general – the<br />

“non-empty” call rate is identical for all<br />

parts of the system, whereas the holding<br />

times are different.<br />

This would be different if the calls waiting<br />

in queue had a limited patience, so<br />

that some of them would leave the queue<br />

without being served. That would reduce<br />

the mean holding time on the source<br />

group and the queue, and the arrival rate<br />

on the server group. Thus the load would<br />

be reduced on all parts of the system.<br />

The linear property of Little’s formula<br />

can be expressed by<br />

A = A1 + A2 = λ1 ⋅ s1 + λ2 ⋅ s2 , (4)<br />

N sources<br />

X X X X<br />

Traffic requests<br />

X X X X X<br />

X X X X<br />

Overflow<br />

Carried calls<br />

n servers<br />

O O<br />

O O O<br />

O O<br />

q waiting positions<br />

Figure 12 Traffic model with overflow from queue<br />

9

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

Saved successfully!

Ooh no, something went wrong!