Traffic Engineering

Traffic Engineering
One billion+ terminals in voice network alone
Plus data, video, fax, finance, etc.

Imagine all users want service simultaneouslyits not even

nearly possible (despite our common intuition)
In practice, the actual amount of equipment provisioned is vastly
less than would support all users simultaneously

And yet, by and large, we get the impression of phone and data
networks that work very well!
How is this possible?
y !!
Traffic theory

Traffic Engineering TradeTrade-offs

Design number of transmission paths, or radio channels?
How many required normally?
What if there is an overload?

Design switching and routing mechanisms

How do we route efficiently?
E.g.
High-usage trunk groups
Overflow trunk groups
g p
Where should traffic flows be combined or kept separate?

Design network topology

Number and sizing of switching nodes and locations
Number and sizing of transmission systems and locations
Survivability

Poisson Distribution
Poisson, is a discrete probability distribution that expresses the
probability of a given number of events occurring in a fixed
i t
interval
l off time
ti
and/or
d/ space if these
th
events
t occur with
ith a known
k
average rate and independently of the time since the last event.
A discrete random variable X is said to have a Poisson
distribution with parameter > 0, if for k = 0, 1, 2, ... the
probability mass function of X is given by:

Where e is the base of the natural logarithm

g
((e = 2.71828...)) k!
is the factorial of k.
The positive real number is equal to the expected value of X
and also to its variance.
variance

Exponential Distribution
In probability theory and statistics, the exponential distribution
(a.k.a. negative exponential distribution) is the probability
di t ib ti
distribution
th t describes
that
d
ib the
th time
ti
b t
between
events
t in
i a Poisson
P i
process, i.e. a process in which events occur continuously and
independently at a constant average rate. It is the continuous
analogue of the geometric distribution,
distribution and it has the key
property of being memoryless.
The probability
d
distribution
b
is

density

function

(pdf)

of

an

exponential

Here > 0 is the parameter of the distribution,

distribution often called the

rate parameter

binomial distribution
In probability theory and statistics, the binomial distribution is
the discrete probability distribution of the number of successes
i a sequence off n independent
in
i d
d t yes/no
/
experiments,
i
t each
h off
which yields success with probability p. Such a success/failure
experiment is also called a Bernoulli experiment or Bernoulli
trial; when n = 1,
1 the binomial distribution is a Bernoulli
distribution.
The binomial distribution is frequently used to model the
number
b
off successes in a sample
l off size n drawn
d
with
h
replacement from a population of size N. If the sampling is
carried out without replacement, the draws are not independent
and so the resulting distribution is a hypergeometric
distribution, not a binomial one. However, for N much larger
than n, the binomial distribution is a good approximation, and
widely used.
used

Probability mass function

In general, if the random variable X follows the binomial distribution
with parameters n and p, we write X ~ B(n, p). The probability of
getting exactly k successes in n trials is given by the probability
mass function:

for k = 0, 1, 2, ..., n, where

is the binomial coefficient,, hence the name of the distribution.

The formula can be understood as follows: we want k successes
(pk) and n k failures (1 p)n k. However, the k successes can
y
among
g the n trials,, and there are different ways
y
occur anywhere
of distributing k successes in a sequence of n trials.

Characterization of Telephone Traffic

Calling Rate () also called arrival rate, or attempts rate, etc.
Average
g number of calls initiated per
p unit time (e.g.
( g attempts
p per
p
hour)
Each call arrival is independent of other calls (we assume)
Call attempt arrivals are random in time
Until otherwise, we assume a large calling group or source pool

If receive

calls from a terminal in time T:

If receive

calls from m terminals in time T:

Group calling rate

g
T

Per terminal
calling rate

m T

Characterization of Telephone Traffic (2)

Calling rate assumption:

Number of calls in time T is
Poisson distributed:

In our case

e x
p ( x)
x!

x 0, 1, 2...

Time
Ti
between
b t
calls
ll is
i -ve

exponentially
ti ll distributed:
di t ib t d

f (t ) e t

0t

mean

Class Question: What do these observations about telephone traffic

imply about the nature of the traffic sources?

-ve Exponential Holding Times

Implies the Memory

Memory--less property
Prob. a call last another minute is independent of how long the call has
already
a
eady lasted!
asted Ca
Call forgets
o gets that
t at itt has
as a
already
eady su
survived
ed to ttime
e T1

PT T1 t T T1 PT t

Proof:

PT T1 t T T1

Recall:

P (T t ) e t / h

P T T1 t T T1

PT T1

P T T1 t
e (T1 t ) / h

T1 / h
PT T1
e
e T1 / h e t / h e t / h PT t

e T1 / h

Characterization of Telephone Traffic (3)

Holding Time (h
h)
Mean length of time a call lasts
Probability of lasting time t or more is also ve exponential in
nature:

P(T t ) e t / h

P(T t ) 0

t0

t0

Real voice calls fits very closely to the negative exponential form
above
As non-voice calls begin to dominate, more and more calls have a
constant holding time characteristic

Departure Rate (
):

1
h

Some Real Holding Time Data

Traffic Volume (V)

V h

= # calls in time period T

h = mean holding
h ldi time
ti
V = volume of calls in time period T

In N. America this is historically usually expressed in terms of centum call seconds

ccs
ccs:
Hundred call seconds

c
c

c
c

s
s

1 ccs is volume of traffic equal to:

one circuit
ci c it b
busy
s fo
for 100 seconds
seconds, or
o
two circuits busy for 50 seconds, or
100 circuits busy for one second, etc.

Traffic Intensity (A)

Also called traffic
traffic flow
flow or simply traffic
traffic.

h
T
T

Recall:
R
ll

Recall:
R
ll
1

R ll
Recall:
V h

= # calls in time period T

h = mean holding time
T = time p
period of observations
= calling rate
= departure rate
V = call volume

Units:

ccs/hour
ccs/hour, or
dimensionless (if h and T are in the same units of time)
Erlang
Erlang unit

The Erlang
Dimensionless unit of traffic intensity
Named after Danish mathematician A. K. Erlang (1878
(1878-1929)
1929)
Usually denoted by symbol E.
1 Erlang
g is equivalent
q
to traffic intensityy that keeps:
p
one circuit busy 100% of the time, or
two circuits busy 50% of the time, or
four circuits busyy 25% of the time,, etc.

26 Erlangs is equivalent to traffic intensity that keeps :

26 circuits busy 100% of the time, or
52 circuits busy 50% of the time,
time or
104 circuits busy 25% of the time, etc.

Class
Could 4 E be produced as a traffic intensity by:
16 sources? (What is the utilization?)
4 sources (same)
1 source?

What is special about the traffic intensity if it pertains to one

source or terminal only?

Erlang (2)
How does the Erlang unit correspond to ccs
ccs?

1 ccs hour

100 call seconds

0.027E
1 hour 60 min hr 60 sec min

3600 call seconds

1E
36 ccs hour
1 hour 60 min hr 60 sec min
Percentage of time a terminal is busy is equivalent to the traffic
generated by that terminal in Erlangs, or
Average number of circuits in a group busy at any time
Typical usages:
residence phone -> 0.02 E
> 0.15
0 15 E
business phone ->
interoffice trunk -> 0.70 E

Traffic Offered, Carried, and Lost

Offered Traffic (TO ) equivalent to Traffic Intensity (A
A)
Takes into account all attempted calls, whether blocked or not,
and
d uses their
h expected
d holding
h ld
times

Also Carried Traffic (TC ) and Lost Traffic (TL )

Consider a group of 150 terminals,
terminals each with 10% utilization (or
in other words, 0.1 E per source) and dedicated service:
service
1

150

each terminal has an

outgoing trunk
(i.e. terminal:trunk ratio = 1:1)

150

TO = A = 150 x 0.10 E = 15.0 E

TC = 150 x 0.10 E = 15.0 E
TL = 0 E

Traffic Offered, Carried, and Lost (2)

A = TO = TC + TL
Traffic
Intensity

Offered
Traffic

Lost
Traffic

Carried
Traffic

TL = TO x Prob
Prob. Blocking (or congestion)
= P(B) x TO = P(B) x A
Circuit Utilization (
) - also called Circuit Efficiency
proportion of time a circuit is busy, or
average proportion of time each circuit in a group is busy

TC

# of Trunks

Grade of Service (gos)

In general, the term used for some traffic design objective
Indicative of customer satisfaction
In systems where blocked calls are cleared, usually use:

gos

TL
TL

P( B)
TO TL + TC

Typical gos objectives:

in busy hour, range from 0.2% to 5% for local calls, however
generally no more that 1%
long distance calls often slightly higher

In systems with queuing, gos often defined as the probability of

d l exceeding
delay
di a specific
ifi length
l
th off time
ti

Grade of Service Related Terms

Busy Hour
One hour period during which traffic volume or call attempts is the
h h
highest
overallll during
d
any given time period
d

Peak (or Daily) Busy Hour

Busyy hour for each day,
y, usuallyy varies from dayy to dayy

Busy Season
3 months (not consecutive) with highest average daily busy hour

High Day Busy Hour (HDBH)

One hour period during busy season with the highest load

Grade of Service Related Terms (2)

Average Busy Season Busy Hour (ABSBH)
One hour period with highest average daily busy hour during the
busy season
busy
For example, assume days shown below make up the busy season:
00:00 to 01:00
01 00 tto 02
01:00
02:00
00
02:00 to 03:00
03:00 to 04:00
04:00 to 05:00
05:00 to 06:00
06:00 to 07:00
07:00 to 08:00
08:00 to 09:00
09:00 to 10:00
10:00 to 11:00
11:00 to 12:00
12:00 to 13:00
13:00 to 14:00
14:00 to 15:00
15:00 to 16:00
16:00 to 17:00
17:00 to 18:00
18:00 to 19:00
19:00 to 20:00
20:00 to 21:00
21:00 to 22:00
22:00 to 23:00
23:00 to 00:00

1-Apr 2-Apr 3-Apr 4-Apr 5-Apr 6-Apr 7-Apr 8-Apr 9-Apr 10-Apr 11-Apr 12-Apr 13-Apr 14-Apr 15-Apr 16-Apr 17-Apr 18-Apr 19-Apr 20-Apr 21-Apr Mean
1.4
1.4
1.2
1.5
1.1
1.5
1.7
1.5
1.0
1.0
1.8
1.5
1.8
1.6
1.2
1.9
1.8
1.6
1.4
1.5
1.2
1.5
12
1.2
18
1.8
16
1.6
13
1.3
10
1.0
16
1.6
11
1.1
11
1.1
10
1.0
12
1.2
17
1.7
20
2.0
20
2.0
18
1.8
13
1.3
17
1.7
14
1.4
19
1.9
11
1.1
14
1.4
15
1.5
15
1.5
1.4
1.8
1.5
1.9
1.2
1.0
1.2
1.1
1.1
1.7
1.5
1.5
1.9
1.9
1.3
1.5
1.8
1.1
1.1
1.2
1.5
1.4
1.2
1.8
1.7
1.4
1.7
1.1
1.5
1.6
1.1
1.9
1.0
1.0
1.4
1.5
1.6
1.1
1.4
1.9
1.4
1.2
1.1
1.4
1.8
1.8
2.3
2.2
2.0
1.7
2.3
1.6
2.2
1.5
2.1
1.6
2.3
2.1
1.7
2.5
1.6
2.0
1.7
1.5
2.3
1.9
2.2
2.3
1.9
2.4
2.5
2.0
2.0
1.7
1.8
1.6
2.0
2.0
2.2
2.2
2.1
1.8
1.6
1.7
2.0
2.3
2.1
2.0
1.7
2.2
1.7
2.5
2.2
2.1
2.2
2.0
2.3
1.6
2.4
2.2
1.5
2.1
2.2
1.8
1.8
1.7
2.1
2.0
2.1
2.0
2.0
2.8
2.2
2.4
2.3
2.4
2.9
2.0
2.4
2.4
2.1
2.9
2.3
2.1
2.9
2.7
2.8
2.3
2.1
2.1
2.7
2.4
3.4
3.1
2.8
2.9
2.5
2.7
2.9
3.0
3.4
3.4
3.1
2.9
2.9
2.9
3.3
3.2
3.5
3.1
3.1
3.1
2.5
3.0
4.0
3.9
3.4
3.5
3.4
3.4
4.0
3.2
3.5
3.4
3.1
3.7
3.3
3.3
3.5
3.9
3.4
3.7
3.7
3.1
3.4
3.9
4.3
4.9
3.8
4.9
5.0
4.7
3.8
5.0
4.8
4.9
4.0
4.9
4.2
4.7
4.2
3.0
4.6
4.4
3.6
4.4
4.1
3.0
5.0
5.0
5.0
4.7
5.0
4.8
4.7
4.3
4.5
3.8
3.4
4.2
4.6
3.2
3.4
4.8
4.1
4.3
4.4
3.6
3.7
4.3
3.8
3.3
4.0
4.2
4.6
4.7
4.0
3.3
3.1
5.0
4.9
4.6
4.1
4.2
3.2
3.6
4.1
4.5
4.2
4.1
4.8
4.6
4.8
4.8
3.2
3.1
4.1
4.5
4.6
4.9
4.7
3.6
3.6
4.8
4.2
4.9
4.4
3.3
3.0
4.2
4.3
4.2
4.7
4.5
4.8
4.7
4.5
4.1
4.4
3.6
3.7
4.5
4.3
4.3
4.9
4.5
3.5
3.5
4.3
4.3
4.3
4.5
4.3
3.3
3.2
4.2
4.4
4.9
4.4
4.8
4.5
3.8
3.2
4.1
4.8
4.4
4.5
4.2
3.3
3.9
4.3
4.9
4.4
4.3
4.5
3.7
3.3
4.2
3.2
3.2
3.8
3.5
3.7
3.1
3.5
3.5
3.2
3.2
3.8
3.4
3.2
4.0
3.3
4.0
3.9
3.0
3.3
3.5
3.3
3.5
Note:
indicates
2.7
2.6
2.7Red2.9
3.3
3.1
3.4
2.9
3.2
2.8
2.7
3.0
3.3
3.2
2.5
2.9
2.8
3.4
3.5
2.9
3.2
3.0
3.0
2.9
3.0
2.7
2.9
3.4
3.3
3.4
2.7
3.3
3.5
3.5
2.7
3.1
3.1
3.3
3.4
3.1
3.0
3.3
3.3
3.1
daily busy hour
3.3
3.3
2.6
3.4
3.2
2.7
2.7
3.4
3.4
3.0
3.0
3.4
3.1
2.8
3.2
3.4
3.0
3.4
3.4
3.1
2.9
3.1
2.9
2.3
2.1
2.9
2.9
3.0
3.0
2.4
2.3
2.9
3.0
2.1
2.2
2.9
3.0
2.6
2.4
2.5
2.7
2.7
2.6
2.6
21
2.1
16
1.6
23
2.3
16
1.6
22
2.2
21
2.1
24
2.4
19
1.9
16
1.6
21
2.1
24
2.4
17
1.7
18
1.8
24
2.4
18
1.8
19
1.9
22
2.2
19
1.9
22
2.2
22
2.2
16
1.6
20
2.0
1.5
2.1
1.9
1.6
1.7
1.6
2.3
2.5
2.4
1.7
2.1
1.8
2.0
2.4
1.7
1.9
2.2
2.3
1.7
2.4
1.8
2.0
1.0
1.1
1.1
1.5
1.8
1.5
1.4
1.8
1.1
1.9
1.2
1.6
1.9
1.8
1.1
1.5
2.0
1.8
1.6
1.4
1.5
1.5

ABSBH

Highest

Hourly Traffic Variations

Daily Traffic Variations

Seasonal Traffic Variations

Seasonal Traffic Variations (2)

Typical Call Attempts Breakdown

Calls Completed - 70.7%
Called Party No Answer - 12.7%
Called Party Busy - 10.1%
Call Abandoned - 2.6%
Dialing Error - 1.6%
Number Changed or Disconnected - 0.4%
Blockage or Failure - 1.9%

3 Types of Blocking Models

Blocked Calls Cleared (BCC
BCC)
Blocked calls leave system and do not return
Good approximation for calls in 1st choice trunk group

Blocked Calls Held (BCH

BCH)
Blocked calls remain in the system for the amount of time it would
have normally stayed for
If a server frees up, the call picks up in the middle and continues
good model of real world behaviour (mathematical
(
Not a g
approximation only)
Tries to approximate call reattempt efforts

Blocked Calls Wait (BCW

BCW)
Blocked calls enter a queue until a server is available
When a server becomes available, the calls holding time begins

Blocked Calls Cleared (BCC)

2 sources
10 minutes
Source #1
Offered Traffic

Source #2
Offered Traffic

1st call arrives and is served

Only one server

2nd call arrives but

server already busy

Traffic
Carried

Total Traffic Offered:

TO = 0.4 E + 0.3 E
TO = 0.7 E

Total Traffic Carried:

TC = 0.5 E

2nd call is cleared

3rd call arrives and is served
4th call arrives and is served

Blocked Calls Held (BCH)

2 sources
10 minutes
Source #1
Offered Traffic

Source #2
Offered Traffic

Total Traffic Offered:

TO = 0.4 E + 0.3 E
TO = 0.7 E

1st call arrives and is served

Only one server

2nd call arrives but server busy

Traffic
Carried

2 2

2nd call is held until server free

2nd call is served

Total Traffic Carried:

TC = 0.6 E

3rd call arrives and is served

4th call arrives and is served

Blocked Calls Wait (BCW)

2 sources
10 minutes
Source #1
Offered Traffic

Source #2
Offered Traffic

Total Traffic Offered:

TO = 0.4 E + 0.3 E
TO = 0.7 E
1st call
ll arrives
i
and
d is
i served
d

Only one server

2nd call arrives but server busy

2nd call waits until server free

Traffic
Carried

Total Traffic Carried:

TC = 0.7 E

2nd call served

3rd call arrives, waits, and
is served
4th call
ll arrives,
i
waits,
i
and
d
is served

Blocking Probabilities
System must be in a Steady State
Also called state of statistical equilibrium
Arrival Rate of new calls equals Departure Rate of
disconnecting calls
Why?
If calls
ll arrive
i faster
f
than
h they
h depart?
d
?
If calls depart faster than they arrive?

Binomial Distribution Model

Assumptions:
m sources
A Erlangs of offered traffic
per source: TO = A/m
probability that a specific source is busy: P(B) = A/m

Can use Binomial Distribution to give the probability that a

certain number (k
k) of those m sources is busy:

m A A
P(k ) 1
k m m
k

mk

A A
m!
1

k!(m k )! m m
k

mk

Binomial Distribution Model (2)

What does it mean if we only have N servers (N<m)?
We can have at most N busy sources at a time
What about the probability of blocking?
All N servers must be busy before we have blocking

P( B) P (k N ) P(k N ) P(k N 1) ... P(k m)

k
mk
m
A A
1
m
k N k m
m

m A A
1 1
m
k 0 k m
N 1

Remember:

m A
P (k )
k m
mk

A
1
m

mk

Binomial Distribution Model (3)

What does it mean if k>N?
Impossible to have more sources busy than servers to serve them
Doesnt accurately represent reality
In reality, P(k>N) = 0

In this model, we still assign P(k>N) = A/m

Acts as good model of real behaviour
Some people call back, some dont

Which type
yp of blocking
g model is the Binomial Distribution?
Blocked Calls Held (BCH)

Time Congestions vs. Call Congestion

Time Congestion
Proportion of time a system is congested (all servers busy)
Probability of blocking from point of view of servers

Call Congestion
Probability that an arriving call is blocked
Probability of blocking from point of view of calls

Why/How are they different?

Time Congestion:

Call Congestion:

P( B) P(k N )

P( B) P(k N )

Probability that all

servers are busy.

Probability that there are

more sources wanting service
th th
than
there are servers.

Poisson Traffic Model

Poisson approximates Binomial with large m and small A/m

e k
P(k )
k!
Note:

= Mean # of
Busy Sources

Poisson lim ( Binomial )

m

What is ?
Mean number of busy sources

=A

e A Ak
P(k )
k!

Poisson Traffic Model (2)

Now we can calculate probability of blocking:

P( B ) P(k N ) P( N ) P( N 1) ... P ()

e A
k!
kN

kN

A A
e
k!

Remember:

e A Ak
P(k )
k!

N 1
1

Ak A
1 e
k 0 k!

Example:

P ( B ) P ( N , A)
P = Poisson

A = Offered Traffic
N = # Servers

P (7,10)
Poisson P(B) with 10 E
offered to 7 servers

Traffic Tables
Consider a 1% chance of blocking in a system with N=10 trunks
How much offered traffic can the system handle?

9
Ak A
Ak A
0.01
e 1 e
k 10 k!
k 0 k!

How do we calculate A?
Very carefully, or
Use traffic
ff tables
bl

Traffic Tables (2)

P(B)=P(N,A)
N
A

Traffic Tables (3)

P(N,A)=0.01

N=10

A=4.14 E

If system with N = 10 trunks

has P(B) = 0.01:
System can handle
Offered traffic (A) = 4.14 E

Poisson Traffic Tables

P(N,A)=0.01

N=10

A=4.14 E

If system with N = 10 trunks

has P(B) = 0.01:
System can handle
Offered traffic (A) = 4.14 E

Efficiency of Large Groups

What if there are N = 100 trunks?
Will they serve A = 10 x 4.14 E = 41.4 E with same P(B) = 1%?
No!
Traffic tables will show that A = 78.2 E!

Why will 10 times trunks serve almost 20 times traffic?

Called efficiency of large groups:
groups

A 4.14

41.4% efficiency
N
10

For N = 10,
10 A = 4.14
4 14 E

For N = 100, A = 78.2 E

A 78.2
78.2% efficiency

N 100

The larger the trunk group, the greater the efficiency

TrafCalc Software
What if we need to calculate P(N,A) and not in traffic table?
TrafCalc
TrafCalc: Custom-designed software
Calculates P(B) or A, or
Creates custom traffic tables

TrafCalc Software (2)

How do we calculate P(32,20)?

TrafCalc Software (3)

How do we calculate A for which P(32,A) = 0.01?

Erlang B Model
More sophisticated model than Binomial or Poisson
Blocked Calls Cleared (BCC)
Good for calls that can reroute to alternate route if blocked
No approximation
pp
for reattempts
p if alternate route blocked too
Derived using birth
birth--death process
See selected pages from Leonard Kleinrock, Queueing Systems
Volume 1: Theory,
Theory John Wiley & Sons
Sons, 1975

Erlang B BirthBirth-Death Process

Consider infinitesimally small time t during which only one
arrival or departure (or none) may occur
Let be the arrival rate from an infinite pool or sources
Let = 1/h be the departure rate per call
Note:
N t if k calls
ll in
i system,
t
departure
d
t
rate
t is
i k

Blockage

Steady State Diagram:

Immediate Service

(N-1)

N-1

Erlang B BirthBirth-Death Process (2)

Steady State (statistical equilibrium)
Rate of arrival is the same as rate of departure
Average rate a system enters a given state is equal to the average
rate at which the system leaves that state
Probability of moving
from state 1 to state 2?

P0

P1

P2
2

P1

PN-1 N-1

(N-1)

Probability of moving
from state 2 to state 1? 2P2

PN N
N

Erlang B BirthBirth-Death Process (3)

P0

P1

P0 P1

P1 P1 2 P2 P0

P1 2 P2

2 P2 P2 3 P3 P1

P2 3 P3

3 P3 P3 4 P4 P2

( N 1) PN 1 PN 1 N PN PN 2
N PN PN 1

Set up balance equations:

P0 P1

Pk 1 k Pk

PN 1 N PN

P2

PN-1

N-1

(N-1)

P1

P0

PN

P
P2
P1 0
2
2
3

P
P3
P2 0
3
6

P
Pk 0
k!

Erlang B BirthBirth-Death Process (4)

Recall:

Rule of Total Probability:

i

P0
P

i

i 0
i 0 i !
N

P0

1
i

i 0 i !
N

For blocking, must be in state k = N:

AN
P( B) B( N , A) PN N !
B = Erlang B
N
N = # Servers
A = Offered Traffic

P
Pk 0
k!

Ai

i 0 i !
N

Recall:
k

Ak 1
k!
Pk k ! N i i
N
A1

i 0 i !
i 0 i !

Ah

Erlang B Traffic Table

B(N,A)=0.001

Example: In a BCC system with

m= sources, we can accept a
0.1% chance of blocking in the
nominal case of 40E offered traffic.
However, in the extreme case of a
20% overload
overload, we can accept a
0.5% chance of blocking.

B(N,A)=0.005

How many outgoing trunks do we

need?

A=40
A 40 E

N=59

Nominal design: 59 trunks

A48 E

Overload design: 64 trunks

Requirement: 64 trunks
N=64

Example (2)
P(N,A)=0.01

N=32

A=20.3 E

P(N,A) & B(N,A) - High Blocking

We recognize that Poisson and Erlang B models are only
approximations but which is better?
Compare them using a 4-trunk group offered A=10E

Erlang B

Poisson

B(4,10)
(4 10) 0.64666
0 64666

P(4,10)
(4 10) 0.98966
0 98966

TC A (1 P ( B)) 10 (1 0.64666)

TC A (1 P ( B)) 10 (1 0.98966)
TC 0.103
0 103E

TC 3.533
3 533E

3.533

0.88
4

0.103

0.026
4

How can 4 trunks handle 10E offered

traffic and be busy only 2.6% of the time?

P(N,A) & B(N,A) - High Blocking (2)

Obviously, the Poisson result is so far off that it is almost
meaningless as an approximation of the example.
4 servers offered enough traffic to keep 10 servers busy full time
(10E) should result in much higher utilization.

Erlang B result is more believable.

All 4 trunks are busy most of the time.

What if we extend the exercise by increasing A?

Erlang B result goes to 4E carried traffic
Poisson result goes to 0E carried

Illustrates the failure of the Poisson model as valid for situations

with high blocking
Poisson only good approximation when low blocking
Use Erlang B if high blocking

Engset Distribution Model

BCC model with small number of sources (m > N)
= mean departure rate per call
= mean arrival rate of a single source
k = arrival rate if in the system
y
is state k

Blockage

k = (m
(m--k)
m

P0

(m-1)

P1

(m-2)

P2

[M-(N-2)]

Immediate Service

[m-(N-1)]

PN-1 N-1

(N-1)

PN N

Engset Distribution Model

Engset's equation is similar to the Erlang-B formula; however it
contains one major difference: Erlang's equation assumes an
i fi it source off calls,
infinite
ll yielding
i ldi
a Poisson
P i
arrival
i l process, while
hil
Engset specifies a finite number of callers. Thus Engset's
equation should be used when the source population is small
(say less than 200 users,
users extensions or customers).
customers)
In practice, like Erlang's equations, Engset's formula requires
recursion to solve for the blocking or congestion probability.
There
h
are severall recursions that
h could
ld be
b used.
d One way to
determine this probability, one first determines an initial
estimate. This initial estimate is substituted into the equation
and the equation then is solved.
solved The answer to this initial
calculation is then substituted back into the equation, resulting
in a new answer which is again substituted. This iterative
process continues until the equation converges to the correct
answer.

Engset Traffic Model (2)

Balance equations give:
k

m!
Pk P0
k !(m k )!

i 0
N

m

i

m
k
Pk i
N
m

i 0 i

therefore:

and

P0

but can show that:

m A

A m

m A N
P( B) P(k N ) E (m, N , A)
i
N
A m

E = Engset
m

A
i
i 0
N

Engset Traffic Table

M = 30 sources

# trunks (N)
Traffic offered (A)
P(B)=E(m,N,A)

N=10

A=4.8 E

Example: 30 terminals each provide

0.16 Erlangs to a concentrator with
a goal of less than 1% blocking.

P(B)<0.01
P(B) 0 01

How manyy outgoing

g g trunks do we
need?
A = 30 x 0.16 = 4.8 E
Requirement: N = 10 Trunks

Check m < 10 x N?
M=30 < 10 x 10 = 100

Erlang C Distribution Model

BCW model with infinite sources (m) and infinite queue length
= arrival rate of new calls
= mean departure rate per call

P0 0

P1 1

P2 2

Immediate Service

Blockage

PQ1 Q1

PQ2 Q2

Erlang C Distribution Model (2)

Balance equations give:
Ak P0
Pk
, kN
k!

and

Ak P0
Pk k N , k N
N N!

P0

and

1
A N N N 1 Ai

N ! N A i 0 i !

But P(B) = P(kN):

P( B)
kN

A P0
P0
A
A
A

P

N
0

N k N N ! k
N!
N ! k 0 N
N N N

AN N
P( B)
P0
N! N A

but can show that:

k

N
A

NA
k 0 N

AN N
C ( N , A) N N ! N AN 1 i
A N
A

N
!
N

i 0 i !
= Erlang
C
l
C

Erlang C Traffic Tables

# trunks (N)

N=18

P(B)=C(N,A)

Traffic offered ((A))

A 7E
A=7

C(18,7)=0.0004
Example:
What
h is the
h probability
b b l off blocking
bl k
in an Erlang C system with 18
servers offered 7 Erlangs of traffic?

Delay in Erlang C
Expected number of calls in the queue?

AN
A
Pk k
(k N ) Pk (k N ) k N P0
N N!
N ! k 0 N
kN
kN

Ak

P0 A N A
N
A C ( N , A)
h

C ( N , A)
N! N A N A
NA
NA
Recall:

Mean #Calls Delayed

y
h
Mean Delay over All Calls =
C ( N , A)

Arrival Rate of Calls

NA
Mean Delay of Delayed Calls =

h
NA

Also:

P (delay T ) C ( N , A)e

h
NA

Comparison of Traffic Models

Erlang C (BCW, sources)
Poisson (BCH, sources)
Erlang B (BCC, sources)
Binomial
i
i l (BCH,
( C m sources))
Engset (BCC, m sources)

P(B)

Offered Traffic (A)

Efficiency of Large Groups

Already seen that for same P(B), increasing servers results in
more than proportional increase in traffic carried
example 1: P (10, 4.14) 0.01

and

P (100, 78.2) 0.01

example
l 2:
2 P (32, 20.3) 0.01

and

P (33, 20.1) 0.005

example 3: B (8, 2.05) 0.001

and

B (80,57.8) 0.001

What does this mean?

If its possible to collect together several diverse sources, you can
provide better gos at same cost, or
provide same gos at cheaper cost

Efficiency of Large Groups (2)

Two trunk groups offered 5 Erlangs each, and B(N,A)=0.002
5E

NN11=13
=?

How many trunks total?

From traffic tables, find B(13,5) 0.002

5E

NN22=13
=?

Ntotal = 13 + 13 = 26 trunks

Trunk efficiency?

TC 10(1 0.002)

0.384
0 384
N
26
38.4% utilization

Efficiency of Large Groups (3)

One trunk group offered 10 Erlangs, and B(N,A)=0.002
How many trunks?
10 E

N=20
N=?

From traffic tables, find B(20,10) 0.002

N = 20 trunks
k

Trunk efficiency?

TC 10(1 0.002)

0.499
N
20

49.9% utilization

For same gos, we can save 6 trunks!

Efficiency of Large Groups (4)

B=0.1

B=0.1

B=0.01

B=0.01

B=0.001

B=0.001

Sensitivity to Overload
Consider 2 cases:
Case 1: N = 10 and B(N,A) = 0.01
B(10,4.5) 0.01, so can carry 4.5 E
What if 20% overload ((5.4 E)?
)

B(10,5.4)
(
) 0.03

3 times P(B) with 20% overload

Case 1: N = 30 and B(N,A) = 0.01
B(30,20.3) 0.01, so can carry 20.3 E
What if 20% overload (24.5 E)?

B(30,24.5) 0.08

8 times P(B) with 20% overload!

Trunk Group Splintering

if high possibility of overloads, small groups may be better

Incremental Traffic Carried by Nth Trunk

If a trunk group is of size N-1, how much extra traffic can it
carry if you add one extra trunk?
Before, can carry: TC1 = A x [1-(B(N-1,A)]
After, can carry: TC2 = A x [1-(B(N,A)]

AN TC 2 TC1 A 1 B( N , A) 1 B( N 1, A)
A B ( N 1, A) B( N , A)
What does this mean?

AN ( N A) B( N , A)
for very low blocking

Random Hunting:
Hunting Increase in trunk groups total carried traffic
after adding an Nth trunk
Sequential Hunting:
Hunting Actual traffic carried by the Nth trunk in the
group

Incremental Traffic Carried by Nth Trunk (3)

Fixed B(N
B(N,A)
A)

AN

Example
Individual trunks are only economic if they can carry 0.4 E or
more. A trunk group of size N=10 is offered 6 E. Will all 10
t k b
trunks
be economical?
i l?

AN A B( N 1, A) B( N , A)
A10 6 B(9, 6) B(10, 6)
6 0.07514 0.04314
0.192 E 0.4 E

At least the 10th trunk is not economical

Binomial Distribution

In probability theory and statistics, the binomial distribution is the discrete

probability distribution of the number of successes in a sequence of n independent
yes/no experiments, each of which yields success with probability p. Such a
success/failure experiment is also called a Bernoulli experiment or Bernoulli trial.
In fact, when n = 1, the binomial distribution is a Bernoulli distribution. The
binomial distribution is the basis for the p
popular
p
binomial test of statistical
significance. A binomial distribution should not be confused with a bimodal
distribution.

Examples

An elementary example is this: Roll a standard die ten times and count the number
of sixes. The distribution of this random number is a binomial distribution with n
= 10 and p = 1/6.

As another example, assume 5% of a very large population to be green-eyed. You

ppick
c 100
00 peop
peoplee randomly.
a do y. Thee number
u be o
of g
green-eyed
ee eyed peop
peoplee you p
pick
c iss a
random variable X which follows a binomial distribution with n = 100 and p =
0.05.

Probability mass function

In general, if the random variable K follows the binomial
distribution with parameters n and p, we write K ~ B(n, p). The
probability
b bilit off getting
tti exactly
tl k successes in
i n trials
t i l is
i given
i
by
b
the probability mass function:

for k = 0, 1, 2, ..., n and where

is the binomial coefficient (hence the name of the distribution)
"n choose k" (also denoted C(n, k) or nCk).
) The formula can be
understood as follows: we want k successes (pk) and n k
failures (1 p)n k. However, the k successes can occur
anywhere among the n trials,
trials and there are C(n, k) different
ways of distributing k successes in a sequence of n trials.

Poisson approximation
The binomial distribution converges towards the Poisson
distribution as the number of trials goes to infinity while the
product
d t np remains
i fixed.
fi d
Therefore the Poisson distribution with parameter = np can be
used as an approximation
pp
to B((n, p) of the binomial distribution
if n is sufficiently large and p is sufficiently small.
According to two rules of thumb, this approximation is good if n
20 and p 0.05,
0 05 or if n 100 and np 10

Poisson distribution
In probability theory and statistics, the Poisson distribution is
a discrete probability distribution that expresses the probability
off a number
b off events
t occurring
i
i a fixed
in
fi d period
i d off time
ti
if
these events occur with a known average rate and
independently of the time since the last event.
The Poisson distribution can also be used for the number of
events in other specified intervals such as distance, area or
volume.

Poisson distribution
The distribution focuses on certain random variables N that
count, among other things, a number of discrete occurrences
(
(sometimes
ti
called
ll d "arrivals")
" i l ") that
th t take
t k place
l
during
d i a timeti
interval of given length. If the expected number of occurrences
in this interval is , then the probability that there are exactly k
occurrences (k being a non-negative
non negative integer
integer, k = 0,
0 1,
1 2,
2 ...)) is
equal to

Poisson distribution
where
e is the base of the natural logarithm (e = 2.71828...)
k is the number of occurrences of an event - the probability of
which is given by the function
k! is the factorial of k
is a positive real number, equal to the expected number of
occurrences that occur during the given interval
interval. For instance
instance, if
the events occur on average 4 times per minute, and you are
interested in the number of events occurring in a 10 minute
interval,, you
y would use as model a Poisson distribution with
= 10*4 = 40.
As a function of k, this is the probability mass function. The
Poisson distribution can be derived as a limiting case of the
binomial distribution.