Traffic Analysis and

Characterization
 ERLANG

Traffic intensity (traffic flow) is expressed in Erlangs.

Erlang =
No. of calls in 1 hour x duration of each call in hours
1 hour
e.g. One call of one hour duration is 1 Erlang
Two calls in 1 hour each of ½ hour duration is 1 Erlang
60 calls in 1 hour each of 1 minute duration is 1 Erlang
Traffic intensity is also expressed in terms of hundred (century) call seconds
per hour (CCS).
1 hour = 3600 seconds=36 hundred seconds
Therefore, 1 Erlang = 36 CCS

2
 ERLANG
The Maximum capacity of a single channel (server) is 1 Erlang
which means it is busy for the whole hour.
The maximum traffic capacity of a network is equal to the
number of channels (servers).
Traffic parameters:
Average Arrival rate = λ per second
Average holding time = t seconds
Traffic intensity A = (λ . t ) Erlangs
Many short calls can produce the same traffic intensity
as a few long calls.

3
 BUSY HOUR

Public Telephone Networks are designed for ‘Busy Hour’

traffic – average activity during the busiest hour of the day.
Residential telephone is typically in use for 5 to 10% of the
busy hour – 0.05 to 0.10 of the busy hour.
This means the traffic intensity of residential telephone is 0.05
to 0.10 Erlangs.
If the average holding time is (say) 3 minutes, the number of
calls per telephone during the busy hour is 1 or 2.

4
 BUSINESS TELEPHONES
Business phones are used more heavily then residential ones.
Busy hour for business traffic is different from residential
traffic.
The Toll (Long distance) trunks are designed to take advantage
of the variations in the traffic pattern.
Two telephones are involved in each connection. The load on the switching
system is one half of the total traffic on the lines connected to the switch.
Call set up time and call release time add to the traffic on trunks.
Offered traffic = carried traffic + lost traffic (loss sys)

5
 ARRIVAL DISTRIBUTIONS
Basic assumption in traffic analysis :

Call arrivals are independent – Arrival from one

source is independent of the arrival from other
sources.
In those cases where call arrivals tend to be
correlate, useful results can be obtained by
modifying a random arrival analysis.
Average call arrival rate from a large group of subscriber
lines=λ (arrivals / second)

6
 ARRIVAL DISTRIBUTION
ASSUMPTIONS

Average call arrival rate from a large group of subscriber lines=

λ (arrivals / second)

1. Only one arrival can occur in any sufficiently small interval

2. The probability of arrival in a small interval ∆t is directly

proportional to the length of the interval ( P  λ .∆t )

3. Probability of an arrival in any particular interval is independent of

what has occurred in other intervals

P (λt) = eλt
This is the probability that ‘t’ seconds elapse from one arrival to the next.

7
 TUTORIAL
Assuming each of the 10000 subscriber lines originates one call per hour,
how often do two calls arrive with less than 0.01 second between them?

8
 TUTORIAL

Average call arrival rate λ = 10000 calls / hour

= 10000/3600 calls per second
= 2.78 arrivals per second
P = e λt λ = 2.78; t = 0.01; λt = 0.0278
P = e 0.0278 = 0.97258 = 0.973
This is the probability that 0.01 seconds elapse from
one arrival to the next. Therefore, (1 – 0.973 =) 0.027 is
the probability of arrivals occurring within 0.01
seconds.
Since 2.78 calls arrive per second, the rate of
occurrence of calls within 0.01 seconds = 2.78 x 0.027 =
0.075 times per second.
 POISSON ARRIVAL DISTRIBUTION

How many arrivals can be expected to occur in some arbitrary

time interval?

Poisson probability
The probability of ‘j’ arrivals in an interval ‘t’
P j (λt) = (λt) j e  λt
j!
When j = 0, P(λt) = e λt

10
 TUTORIAL

What is the probability that a 1000-bit data block experiences exactly four
errors while being transmitted over a transmission link with a bit error
rate of (BER) of 10 -5 ? Assume independent errors.

The average number of errors λt = 1000 x 10 -5 = 0.01

Prob (4 errors) = P4 (0.01) = (0.01)4 e0.01

4!
= 4.1667 x 10 -10 x 9.9 x 10 -1
= 4.125 x 10 -10

11
 TUTORIAL
• Given a message-switching node that normally experiences four arrivals
per minute, what is the probability that eight or more arrivals occur in an
arbitrarily chosen 30-sec interval?

• The average number of arrivals in a 30 sec interval is

» λt=4 X 30/60 = 2
• The probability of eight or more arrivals (when the average is 2) is
» 𝑃≥8 = ∞ 𝑖=8 𝑃𝑖 (2)
7
=1 - 𝑖=0 𝑃𝑖 (2)
21 22 23 27
=1 - 𝑒 −2 1+ + + + …….+
1! 2! 3! 7!
= 0.0011
 HOLDING TIME DISTRIBUTIONS

A = λt λ: average arrival rate; t: holding time;

Two cases considered constant holding time
exponential holding time
Constant holding time :
Probability of j circuits being busy at any particular time
(when all the requests are serviced) is given by Poisson
formula. Pj (λt)= Pj (A)= Aj eA
j!
2. Exponential holding time - more complex
Same Poisson formula holds good.

13
 DIMINISHING RETURNS
• Assume that a trunk group has enough channels to immediately carry all
of the traffic offered to it by a Poisson process with an arrival rate of one
call per minute. Assume that the average holding time is 2 min. What
percentage of the total traffic is carried by the first five circuits, and how
much traffic is carried by all remaining circuits? ( Assume that the traffic is
always packed into the lowest numbered circuits.)
 DIMINISHING RETURNS

Arrival rate = 1 call per minute

Av. Holding time= 2 minutes
Total offered traffic = 2 Erlangs
Assume sufficient circuits are there to carry this traffic.
What percentage of this traffic is carried by the first 5 ccts?

It can be shown that the traffic carried by the first 5 circuits is 1.89
Erlangs (94.5%). The remaining circuits carry only 0.11Erlangs (5.5%).
If only 5 circuits are provided, 5.5% of the traffic will be lost.
For 2 Erlang offered traffic, if only 5 circuits are provided, the blocking
probability is 5.5%.

15
 Lost calls cleared
• The blocked calls in one trunk group are cleared via another
• trunk group.
• LCC is well suited to study the behavior of trunk
transmission system.
• Main purpose of study is to estimate blocking probability.
• Blocking probability is calculated using Erlang's formulation.
• Based on the principle of STATISTICAL EQUILIBRIUM.
• In equilibrium a system is as likely to have an arrival as it is
to have a termination.
 Lost calls cleared
• Erlang formula is based on the assumption that, call
arrivals
• are independent of no. of active callers.
• It is justified when M >>>> N, (not always possible, for
• example a 3 stage space division network)
• If considering finite no. of subscribers, arrival rate
depends on subscribers who are not occupied as busy
and do not generate new calls.
• Traffic in this case is 'ENGSET TRAFFIC'
• In some books it is referred as ENGEST TRAFFIC
 ERLANG –B FORMULA
For lost calls cleared case

Blocking probability : B
Offered traffic intensity : A
Number of circuits provided : N
AN
B= N
N! ∑i=0 (Ai / i !)
Erlang B formula is more accurate than the Poisson
formula and is adopted by ITU and the industry.

18
 PLOT OF ERLANG- B FORMULA
LOST CALLS CLEARED SYSTEM
1.0
0.5
B N=1
0.1
N=2
0.05
3
4
0.01 5

0.005
6

0.001

50

0.0001
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
OFFERED TRAFFIC PER CHANNEL (Erlangs) 19
 ERLANG B TRAFFIC TABLES
Maximum offered load 'A' Erlangs for various blocking probabilities 'B' and number
of servers (circuits/channels) 'N'
BLOCKING PROBABILITY B (%)

0.05 0.01 0.5 1.0 2.0 5.0 10.0 15.0 20.0 30.0 40.0
N
1 0.0001 0.0005 0.001 0.005 0.010 0.020 0.053 0.111 0.176 0.250 0.429 0.667
2 0.014 0.032 0.046 0.105 0.153 0.223 0.381 0.595 0.796 1.00 1.45 2.00

3 0.087 0.152 0.194 0.340 0.455 0.602 0.899 1.27 1.60 1.93 2.63 3.48
4 0.235 0.362 0.439 0.701 0.869 1.09 1.62 2.05 2.50 2.95 3.89 5.02
5 0.452 0.649 0.762 1.13 1.36 1.66 2.22 2.88 3.45 4.01 5.10 6.60
6 0.728 0.996 1.15 1.62 1.91 2.28 2.96 3.76 4.44 5.11 6.51 8.19
7 1.05 1.39 1.58 2.16 2.50 2.94 3.74 4.67 5.46 6.23 7.86 9.80
8 1.42 1.83 2.05 2.73 3.13 3.63 4.54 5.60 6.50 7.37 9.21 11.40
9 1.83 2.30 2.56 3.33 3.78 4.34 5.37 6.55 7.55 8.52 10.60 13.00
10 2.26 2.80 3.09 3.96 4.46 5.08 6.22 7.51 8.62 9.68 12.00 14.70
11 2.72 3.33 3.65 4.61 5.16 5.84 7.08 8.49 9.69 10.90 13.30 16.30
12 3.21 3.88 4.23 5.28 5.88 6.61 7.95 9.47 10.80 12.00 14.70 18.00
13 3.71 4.45 4.83 5.96 6.61 7.40 8.83 10.50 11.90 13.20 16.10 19.60
14 4.24 5.03 5.45 6.66 7.35 8.20 9.73 11.50 13.00 14.40 17.50 21.20
15 4.78 5.63 6.08 7.38 8.11 9.01 10.60 12.50 14.10 15.60 18.90 22.90
20
 N OFFERED TRAFFIC ‘A’ ERLANGS

16 5.34 6.25 6.72 8.10 8.88 9.83 11.50 13.50 15.20 16.80 20.30 24.50
17 5.91 6.88 7.38 8.83 9.65 10.70 12.50 14.50 16.30 18.00 21.70 26.20
18 6.50 7.52 8.05 9.58 10.40 11.50 13.40 15.50 17.40 19.20 23.10 27.80
19 7.09 8.17 8.72 10.30 11.20 12.30 14.30 16.60 18.50 20.40 24.50 29.50
20 7.70 8.83 9.41 11.10 12.00 13.20 15.20 17.60 19.60 21.60 25.90 31.20
21 8.32 9.50 10.10 11.90 12.80 14.00 16.20 18.70 20.80 22.80 27.30 32.80
22 8.95 10.20 10.80 12.60 13.70 14.90 17.10 19.70 21.90 24.10 28.70 34.50
23 9.58 10.90 11.50 13.40 14.50 15.80 18.10 20.70 23.00 25.30 30.10 36.10
24 10.20 11.60 12.20 14.20 15.30 16.60 19.00 21.80 24.20 26.50 31.60 37.80
25 10.90 12.30 13.00 15.00 16.10 17.50 20.00 22.80 25.30 27.70 33.00 39.40
26 11.50 13.00 13.70 15.80 17.00 18.40 20.90 23.90 26.40 28.90 34.40 41.10
27 12.20 13.70 14.40 16.60 17.80 19.30 21.90 24.90 27.60 30.20 35.80 42.80
28 12.90 14.40 15.20 17.40 18.60 20.20 22.90 26.00 28.70 31.40 37.20 44.40
29 13.60 15.10 15.90 18.20 19.50 21.00 23.80 27.10 29.90 32.60 38.60 46.10
30 14.20 15.90 16.70 19.00 20.30 21.90 24.80 28.10 31.00 33.80 40.00 47.70
B% 0.01 0.05 0.1 0.5 1.0 2.0 5.0 10.0 15.0 20.0 30.0 40.0

21
 EXAMPLE

• For example, if a group of user made 30 calls in one hour, and

each call had an average call duration of 5 minutes, then the
number of Erlangs this represents is worked out as follows:
• Minutes of traffic in the hour=number of calls x duration
Minutes of traffic in the hour=30 x 5
• Minutes of traffic in the hour=150
• Hours of traffic in the hour=150 / 60
• Hours of traffic in the hour=2.5
• Traffic figure=2.5 Erlangs

22
 TRUNK GROUP EXAMPLE
1 4 CLUSTERS – EACH WITH 22 TMLS
1 EACH TERMINAL ACTIVE FOR 10% OF TIME
22 BLOCKING PROB:5% 1

2
22
CP . C
U
. P
.
1 U
.
3 .
22 .
.
1

4
22 88

88 x 0.1 = 8.8 Erl

(22 x 0.1) =2.2 Erl
Requires 13 circuits
Requires 5 circuits
Total= 4 x 5 = 20 ccts

23
 TRUNK GROUPS
SMALLER vs LARGER

• Large trunk groups are more efficient than small ones

• Excess trafic in one group can use idle circuits in another
group
This leads to Hierarchical switching structures.

sw

MESH CONNECTED STAR CONNECTED

PRACTICAL NETWORK HAS A MIX OF BOTH

24
 NATIONAL LONG DISTANCE NETWORK
HIERARCHY OF SWITCHES

TAX Level I 22 SWITCHES

TAXLevel II 322 SWITCHES

Level I switches are mesh connected.

Other switches are mesh + star connected.
Exchanges are connected to the TAX through star network.

25
 LOST CALLS RETURNING
Lost calls cleared : unserviceable (blocked) calls leave the
system and never return
Lost calls returning: blocked calls return to the system
as retries.
Assumptions in analysis:
1. All lost calls return to the system and eventually get
serviced (sometimes with multiple retries)
2. Elapsed time between blocking and retries are random and
statistically independent of each other
3. Waiting time before the occurrence of retries is somewhat
longer than the average holding time of the
connection.

26
• A T1 line is used to carry traffic from a remote
concentrator to a central office. How many 10
CCS subscribers can the concentrator system
support at 0.5% blocking. Compare an infinite
source analysis to a finite source analysis.
Assume blocked calls cleared
 LOST CALLS RETURNING -ANALYSIS
Consider a system with a call arrival rate of λ.
If a percentage B of the calls is blocked, B times λ
retries will occur in the future. Of these, again B percent
will get blocked and Bλ retries will occur. Continuing in
this manner the total arrival rate λ’ can be expressed as
λ’= λ + Bλ + B ² λ + B ³ λ + …………….
= λ / (1 B)
B is the blocking probability from lost calls cleared analysis with
traffic intensity A’ = λ’ .t.
First determine the estimate of B using λ and then calculate λ’ .
Next use λ’ to obtain a new value of B and an updated value of λ’ .
Continue this until values of λ’ and B are obtained.
28
 Tutorial
• What is the blocking probability of a PBX to a central
office trunk group with 10 circuits servicing a first-
attempt offered traffic load of 7 erlangs? What is the
blocking probability if the number of circuits is
increased to 13? Assume random retries for all
blocked calls.
 BLOCKING PROBABILITY OF LOST CALLS RETURNING

1.0

B
0.5

1
0.1

0.05

LOST CALLS CLEARED

10
.001

20

0.0001
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

A: OFFERRED TFC PER CHANNEL 30

 LOST CALLS HELD

This is a theoretical case in which the blocked calls are held by

the system and serviced when facilities become available.
Switched telephone network does not operate on the ‘Lost
calls Held’ mode.
This analysis is applicable in TASI system.

This is different from the Delay systems.

31
 LOST CALLS HELD
• A TASI (Time Assignment Speech Interpolation) system concentrates some number of voice
sources onto a smaller number of transmission channels.
• A source receives service (is connected to a channel) only when it is active.
• If a source becomes active when all channels are busy, it is blocked and speech clipping
occurs.
• Each speech segment starts and stops independently of whether it is serviced or not.
• TASI systems were originally used on analog long-distance transmission links such as
undersea cables.
• More modern counterparts of TASI are referred to as digital circuit multiplication (DCM)
systems.
• In contrast to the original TASI systems, DCM systems can delay speech for a small amount
of time, when necessary, to minimize the clipping.
• In this case, a lost calls held analysis is not rigorously justified because the total time a
speech segment is "in the system" increases as the delay for service increases.
• However, if the average delay is a small percentage of the holding time, or if the coding rate
of delayed speech is reduced to allow the transmission channel time to "catch up," a lost
calls held analysis is still justified.
 Tutorial
• A TASI system has 10 channels and 20 sources
connected to it. What is the Probability of
clipping if the activity factor for each source is
0.4.
 Tutorial
• What is the probability that a talk spurt
experiences clipping in a TASI system with 10
sources and 5 channels? With 100 sources and
50 channels? Assume that the activity factor
of each talker is 0.4.
 LOST CALLS CLEARED (LCC) SYSTEM for Finite
sources
• Erlang formula is based on the assumption that,
call arrivals are independent of no. of active
callers.
• It is justified when M >>>> N, (not always
possible, for example a 3 stage space division
network)
• If considering finite no. of subscribers, arrival
rate depends on subscribers who are not
occupied as busy and do not generate new calls.
• Traffic in this case is 'ENGSET TRAFFIC'
 LOST CALLS CLEARED (LCC) SYSTEM for Finite
sources
• When considering finite sources, one more parameter
time congestion has to be considered.
• Time congestion is the percentage of time that all
servers in a group are busy.
• It is identical to the probability that all servers are busy at
randomly selected times. However, time congestion is
not necessarily identical to blocking probability( which is
sometimes referred to as call congestion)
• Time congestion merely specifies the probability that all
servers are busy. Before blocking can occur, there must
be an arrival.
 LOST CALLS CLEARED (LCC) SYSTEM for Finite
sources
• In an infinite sources system time congestion and call congestion are
identical because the percentage of arrivals encountering all servers
busy is exactly equal to the time congestion.
• In a finite source system however, the percentage of arrivals
encountering congestion is smaller because fewer arrivals occur during
periods when all servers are busy.
• Thus in a finite source system call congestion(blocking probability)
• is always less than the time congestion.
• As an extreme example consider equal numbers of sources and servers
The time congestion is the probability that all servers are busy.
• The blocking probability is obviously zero
LOST CALLS CLEARED (LCC) SYSTEM for Finite
sources
LOST CALLS CLEARED (LCC) SYSTEM for Finite
sources
 DELAY SYSTEMS
• Also referred as Waiting call systems or Queuing systems
• Systems that delay non serviceable requests until necessary
facilities become available
• Analysed using 'Queuing Theory' also known as waiting line
theory.
• Waiting calls are usually processed on First come First serve
basis.
• Applications in:
– Message switching
– Packet switching
– Digit receivers
– Automatic call distribution, etc
 DELAY SYSTEMS
• While analysing Queuing system, we come across a no.
of random variables such as:
– No. of waiting requests
– Inter-arrival times between request
– Queue length
– Time spent by a request in the system and many more

• A delay system allows for greater utilization of servers

than a loss system.
• Although the arrivals to system are random, the
servers see a somewhat regular arrival pattern.
 DELAY SYSTEMS
• Ideally there is a need of infinite queuing
capacity, since no limit on no. of arrivals.
• Practically Queues with finite capacity is only
possible.
• Therefore, there is a probability (very small) of
blocking in delay systems.
• Two characteristics of grade of service must be
considered-
1. Delay Probability
2. Blocking Probability
 DELAY SYSTEMS
• Assuming that a delay system has finite queue
capacity, a necessary condition for its stable
operation is -
– (Mean arrival rate/Mean service rate)< 1
– or
– (Offered Traffic/No. of servers)< 1
• If these conditions are not satisfied the queue
length would become infinite sooner or later
 DELAY SYSTEMS

• For Example: M/M/1 = Single server system with

random input and negative exponential service
time.
 DELAY SYSTEMS
• In case of delay systems :
– Total time, a request is in the system= waiting time+
holding time
• Holding time = = Service time
• In contrast to loss system, delay system performance is
generally dependent on the distribution of service
times and not just the mean value of tm
• We need to study two cases:
– Exponential Service Times
– Constant Service Times
 DELAY SYSTEMS
• Exponential Service Times
• systems represented as M/M/N
• Assumptions: Calls are serviced in the order of their arrivals
• Assumptions: Probability of an arrival is independent of the no. of request
already in queue
• So, the probability that a call experiences congestion and therefore it is
delayed is given by-

• Where, N= no. of servers, A= ofered load and B = Blocking Probability for

LCC system.
• This is Erlang Second Formula E2N(A) or Erlang's Delay Formula or Erlang's C
formula
• For a single server system , N=1 (M/M/1):
• Prob(delay) = Output utilization traffic carried by the server= ρ
 DELAY SYSTEMS
• Exponential Service Times
• Probability that a call arrival at any randomly chosen instant is
delayed for more than t service times or Probability that waiting time
exceeds 't'-

• Average waiting time for all arrivals or expected delay for all arrivals
(t)-

• Average delay of only those arrivals that are delayed is -

 DELAY SYSTEMS
• Constant Service Times
• These are delay systems with random arrivals, constant service
• times and a single server : M/D/1
• Assumption: FIFO Service discipline
• Infinite Sources
• The average waiting time for a single server with constant service
time is –

• NOTE : In both the cases (constant and exponential service times)

Server Utilization = ρ = A
 DELAY SYSTEMS
QUEUING THERY

QUEUING SYSTEM, WAITING SYSTEM

Call arrivals occurring when all servers are busy are placed
in queue and held until service commences.
This analysis is applicable to several services other
than telecom network traffic –
taxis, buses, cash counters, airport landings, inventory
control, etc.
Telecom applications are – Message switching, Packet
Switching, Statistical Multiplexers, ACD, PBXs.

50