You are on page 1of 28

TELETRAFFIC ENGINEERING

TELETRAFFIC
THEORY

Teletraffic Theory
Teletraffic theory - the application of probability disciplines (stochastic
processes, queueing theory, numerical simulation) to:

planning,
performance evaluation,
operation,
maintenance

of telecommunication systems.
The objective:

to make the traffic measurable through mathematical models,


to derive the relationship between QoS and system capacity.

The task:

to design systems as cost effectively as possible with a predefined QoS


(assuming the future traffic demand and the capacity of system elements).
to specify methods for controlling QoS and emergency actions when
systems are overloaded or technical faults occur.

Teletraffic Theory
Teletraffic theory is an inductive discipline:
From observations of real systems we
establish theoretical models, from which
we derive parameters, which can be
compared with corresponding
observations from the real system.
If there is agreement, the model has been
validated. If not, then we have to
elaborate the model further.
This scientific way of working is called the
research spiral.
3

Traffic Concept

Traffic in Circuit-Switched
Networks
We usually use the word traffic to denote the traffic
intensity
Instantaneous traffic intensity (sometimes called
occupancy): measure of the occupancy of a servers or resources (the
number) at the time t, [Erlang]
Average traffic intensity:
t

1 2
Y t1, t2
n t dt ,
t2 - t1 t`
where n t

- the number of occupied servers at the time t

Note! Traffic intensity is usually referred to as average traffic intensity5

Traffic in Circuit-Switched
Networks
Carried traffic: traffic carried by the group of servers during
the time T, AC
Note! The carried traffic cannot exceed the number of channels, single channel can at
most carry 1 Erlang of the traffic!

The carried trac


(= number of busy
devices)
as a function n(t) of t.

We use the average traffic


intensity during a period of
time T (mean).
6

Traffic in Circuit-Switched
Networks
Traffic volume: measure of the total traffic carried in a period of
time or (total work done by a resource) ,[Erlang-hours]
Note! Losses are caused by inability to carry all traffic

Offered traffic: traffic which would be carried if no arrivals were rejected


due to the lack of capacity (theoretical, cannot be measured).
, where l - average number of arrivals per time unit,
- average service time of arrival.
Lost traffic (rejected, blocked):. AL A - AC
Note! the value of lost traffic is reduced by increasing the capacity of the
system (when the capacity of the system tends to infinity AC A

Example: arrival intensity is 10 arrs/m, average service time 2


minutes A 10 2 20 [Erlangs]
7

Two main types of traffic are defined: offered and carried

Traffic Variations
Traffic in circuit-switched networks varies according to
activity of users:
Traffic is generated by single sources - subscribers,
subscribers are assumed to be independent.

Traffic is characterized by two major components:


Stochastic variations:
Random generation of calls by subscribers.
Deterministic component:
nearly deterministic variability of number of calls over days,
weeks, months and even years (subscribers needs to make
more calls in a certain period of time).

Variations in traffic can be also split up into:


Variations in service times,
Variations in call intensity (number of calls).

Traffic Variations

Number of calls per minute to a switching center


9

Traffic Variations

10

Average number of voice calls per minute: 10 workdays averages

Traffic Variations

11

Average service times for voice calls

Traffic Variations
Traffic patterns looks very similar for different days:

Traffic patterns are similar during week-days,


Traffic patterns are similar during week-end days,
Traffic patterns are different during week-days and week-end
days.

The deterministic variations in the traffic can be divided


to:

24 hours variations
Weekly variations: highest trac: Monday, then on Friday,
Tuesday, Wednesday, Thursday, Saturday, Sunday.
Year variations: for example: very low traffic in vacation times
(July in Finland).
Very large-scale variation: traffic increases depending on
12
technology development and economics.

Traffic Variations

13

Mean holding time for trunk lines as a function of time of day

Traffic variations

14

Average service times for modem calls

Traffic variations

15

Average number of calls per 24 hours to a switching center

Busy Hour
Questions:
When the peak number of calls occurs?
Is this peak the same for each day?
Note! The highest traffic does not occur at same time every day.

Time consistent busy hour (TCBH) is defined as those 60


minutes (determined with an accuracy of 15 minutes) which during a
long period on the average has the highest traffic.
Note! It happens, the traffic during the busiest hour is larger than the TCBH,
but on the average over several days it will be the largest.
Note!
We considered a traditional PSTN traffic
Other traffic types or other circuit-switched networks have their own
patterns and variations:

Dial-up Internet via modems,


Voice calls in GSM/UMTS mobile networks.

16

Blocking Concept
The PSTN is dimensioned so that all subscribers cannot
be connected at the same time.
Usual dimensioning rules applied:

~ 5 8% should be able to make domestic calls at the same


time,
~ 1% should be able to make international calls at the same
time.

There should be some problems:

Resources are shared with many others,


It is possible that a subscriber cannot establish a call.

When it is not possible to establish a call it:

Has to wait,
Has to be blocked.

17

Blocking Concept
Depending on how system operates we
distinguish between:
Loss systems: arrival is lost when there are
insufficient resources in the system,
Waiting systems: arrival waits when there are
insufficient resources in the system,
Mixed loss-waiting systems: depending on arrival
it can wait of get lost (the number of waiting
positions - buffer - is limited).

18

Blocking Concept
Networks performance measures can be
expressed using:
Call congestion: fraction of call attempts that
observes all servers busy (the user-perceived
quality-of-service),
Time congestion: fraction of time when all servers
are busy,
Traffic congestion: the fraction of the offered traffic
that is not carried, possibly despite several
attempts,
Waiting time: mean and distribution in delay
systems (queueing systems)
19

Simple Teletraffic Model

20

10

Infinite System

21

Pure Loss System

22

11

Poisson Model

23

State Transition Diagram

24

12

Equilibrium Distribution

25

Equilibrium Distribution

26

13

Erlang Model

27

State Transition Diagram

28

14

Equilibrium Distribution

29

Equilibrium Distribution

30

15

Time Blocking

31

Call Blocking

32

16

Application to Telephone Traffic


Modeling in Trunk Network

33

Binomial Model

34

17

On-Off Type Customer

35

On-Off Type Customer

36

18

State Diagram Transition

37

Equilibrium Distribution

38

19

Equilibrium Distribution

39

Engset Model

40

20

State Transition Diagram

41

Equilibrium Distribution

42

21

Equilibrium Distribution

43

Time Blocking

44

22

Call Blocking

45

Call Blocking

46

23

Call Blocking

47

Call Blocking

48

24

Application to Telephone Traffic


Modeling in Access Network

49

Erlang-C Model

50

25

QoS in Lossless Systems of


M/M/C Type
M/M/C systems do not reject service
requests
If the resources are not available the
request is placed in the queue
Queue is assumed infinite
GOS is inadequate measure of QoS
Some examples
dispatch voice

low end packet data services (SMS,


MMS, e-mail, )

Relevant QoS parameters in lossless


systems
Probability of service delay

Average delay for all requests

Average delay for the requests


placed in the queue

90% delay percentile

Average number of requests in the


queue

Probability of a delay that exceeds


a given threshold

51

Trunking Model for Lossless


Systems - Erlang C

Erlang C assumptions

Call arrival process is


Poisson
Service time is
exponentially distributed
There are C identical
servers (channels)
The queue is infinite

Common QoS
parameter in M/M/C
systems are

delay probability
average delay
number of users in the
queue

State diagram of M/M/C system


l

2m

C-1

(c - 1) m

l
C

cm

C+1

cm

Probability of delay

aC
C!1 -
EC a, C C -1 n
a
aC

C!1 -
n 0 n!

l
m

a
Server utilization
C

Erlang C delay formula

Offered traffic
52

26

Erlang C - Performance Curves

Erlang C family of curves

Erlang C formula
can be given in a
form of

-1

family of curves
table (Appendix)

10

probability of delay

C = 1,2,...,14;15,20,...,30;40,50,...,100;

-2

10

-3

10 -1
10

10

traffic [erlangs]

10

10
53

Erlang C - Summary of
Performance Parameters
Quantity
Offered traffic

Symbol
a

Channel utilization

Probability of having no calls


at the cell site

p0

C -1 a k

aC

k 0 k! C!1 -

pn

an
n! p 0
C
a n- C p 0
C!

Probability of having exactly n


calls at the cell site

Probability of service request


delay

Pr 0

Average number of requests in


the queue

Lq

Average time that all call


requests spend in the queue
Average time that delay call
request spend in the queue

D1

D2

Value (Formula)
al m
a

C
-1

,n C
,n C

aC
C!1 -
C -1
an
aC

C! 1 -
n 0 n!
a
Pr 0
C-a
Pr 0

Alternative notation
Pr 0 E a, C

Probability of delay
exceeding T1
T

P t T1 P t 0 exp - C - a 1
H

H average call holding


time

H
C-a

H
C-a

54

27

Erlang C - Examples
Example. Consider a cell site supporting MMS service. Assume that the messages
are exponentially distributed with average length of 2 Meg. The cell site provides
two channels that have transfer rate of 200kbps. If there are 125 requests per
hour estimate probability of
1)
Request being delayed
2)
Request being delayed by more than 10 sec
Answers:

Call holding time: 10 sec


Average service rate: 0.1 request/sec

Birth rate: 0.034 requests/sec

Offered traffic: 0.034 E

Average resource utilization: 0.17


Probability of delay: 0.04940 ~ 5%
1)
Probability of delay exceeding 10 sec: 0.009 ~ 1%
2)
55

28

You might also like