Chapter 6

Traffic
Traffic Engineering
Engineering
Typical Traffic Distribution
on a Cellular System
80%
100%
90% SUN
Efficiency %
80%
MON
41
70%
60% TUE

50% WED
40%
30%
THU Capacity,
20%
FRI
Erlangs
10% SAT
0%
1 # Trunks 50
Hour

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-1

 A Game of Avoiding Extremes

The traffic engineer must walk a fine line

between two problems:
■ Overdimensioning
• too much cost
• insufficient resources to construct
• traffic revenue is too low to
support costs
• very poor economic efficiency!
■ Underdimensioning
• blocking
• poor technical performance
(interference)
• capacity for billable revenue is low
• revenue is low due to poor quality
• users unhappy, cancel service
• very poor economic efficiency!

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-2

 Dimensioning the System:
An Interactive, Iterative Process
■ Some traffic engineering decisions trigger
resource acquisition
• additional blocks of numbers from the
local exchange carrier PSTN
Office
• additional cards for various functions in
the switch and peripherals Cell
• additional members in PSTN trunk DMS-MTX
groups; additional T-1/E-1s to busy sites
■ Some traffic engineering decisions trigger
more engineering
• finding more frequencies to add to
blocking sites
• adding additional cells to relieve blocking
• finding short-term fixes for unanticipated
problems
■ This course is concerned primarily with
determining the number of voice channels
required in cells, with the related site
engineering and frequency or code planning

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-3

 Basics of Traffic Engineering
Terminology & Concept of a Trunk

■ Traffic engineering in telephony is focused on the voice paths

which users occupy. They are called by many different names:
• trunks
• circuits
• radios, transceivers (“TRXs”), channel elements (CDMA)
■ Some other common terms are:
• trunk group
– a trunk group is several trunks going to the same
destination, combined and addressed in switch
translations as a unit , for traffic routing purposes
• member
– one of the trunks in a trunk group

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-4

 Units of Traffic Measurement

Traffic is expressed in units of Circuit Time

General understanding of telephone traffic engineering began

around 1910. An engineer in the Danish telephone system,
Anger K. Erlang, was one of the first to master the science of
trunk dimensioning and publish the knowledge for others. In his
honor, the basic unit of traffic is named the Erlang.
■ An Erlang of traffic is one circuit continuously used during an
observation period one hour long.
Other units have become popular among various users:
■ CCS (Hundred-Call-Seconds)
■ MOU (Minutes Of Use)
■ It’s easy to convert between traffic units if the need arises:
1 Erlang = 60 MOU = 36 CCS

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-5

 How Much Traffic Can One Trunk Carry?
■ Traffic studies are usually for periods of one hour
■ In one hour, one trunk can carry one hour of traffic -- One Erlang
■ If nothing else matters, this is the limit!
■ If anyone else wants to talk -- sorry!

Absolute Maximum Capacity

of One Trunk
y!
One Trunk
rr
So

Constant One Erlang

Talker

■ We must not plan to keep trunks busy all the time. There must be
a reserve to accommodate new talkers! How much reserve? next!
December, 1997 CDMA BSM, BSC, BTS 6-6
 Traffic Engineering And Queuing Theory

Ticket counter analogy ■ Traffic engineering is an application of a

Servers science called queuing theory
• Queuing theory relates user arrival
statistics, number of servers, and
various queue strategies, with the
probability of a user receiving service
Queue
• If waiting is not allowed, and a blocked
call simply goes away, Erlang-B
User population formula applies (popular in wireless)
• If unlimited waiting is allowed before a
call get served, the Erlang-C formula
Queues we face in applies
everyday life • If a wait is allowed but is limited in
1) for telephone calls time, Binomial & Poisson formulae
2) at the bank apply
3) at the gas station • Engset formulae apply to rapid,
4) at the airline counter packet-like transactions such as
paging channels

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-7

 Offered And Carried Traffic
PSTN or other ■ Offered traffic is what users attempt to
Wireless user originate
■ Carried traffic is the traffic actually
successfully handled by the system
■ Blocked traffic is the traffic that could
Carried not be handled
Traffic • Since blocked call attempts never
materialize, blocked traffic must be
MTXBSC estimated based on number of
blocked attempts and average
duration of successful calls
BTS BTS BTS BTS BTS BTSBlocked

Traffic Offered Traffic =

Carried Traffic + Blocked Traffic
Offered
Traffic
TOff = NCA x TCD
TOff = Offered traffic
NCA = Number of call attempts
TCD = Average call duration
October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-8
 Principles of Traffic Engineering
Blocking Probability / Grade of Service
■ Blocking is inability to get a circuit when one is needed
■ Probability of Blocking is the likelihood that blocking will
happen
■ In principle, blocking can occur anywhere in a wireless system:
• not enough radios, the cell is full
• not enough paths between cell site and switch
• not enough paths through the switching complex
• not enough trunks from switch to PSTN
■ Blocking probability is usually Typical Wireless System
expressed as a percentage Design Blocking Probabilities
using a “shorthand” notation: PSTN Office
• P.02 is 2% probability, etc. Cell

• Blocking probability sometimes Cell

is called “Grade Of Service” P.005

DMS-MTX Cell
■ Most blocking in cellular systems
occurs at the radio level. P.02
• P.02 is a common goal at the P.001 P.005
radio level in a system
October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6-9
 Number of Trunks
vs. Utilization Efficiency
■ Imagine a cell site with just one voice channel. At a P.02
Grade of Service, how much traffic could it carry?
• The trunk can only be used 2% of the time, otherwise the
blocking will be worse than 2%.
• 98% availability forces 98% idleness. It can only carry
.02 Erlangs. Efficiency 2%!

■ Adding just one trunk relieves things greatly. Erlang-B P.02 GOS
Now we can use trunk 1 heavily, with trunk 2 Trks Erl Eff%
handling the overflow. Efficiency rises to 11% 1 0.02 2%
2 0.22 11%

The Principle of Trunking Efficiency 80%

■ For a given grade of service, trunk Efficiency %
41
utilization efficiency increases as the
number of trunks in the pool grows larger.
• For trunk groups of several hundred, Capacity,
utilization approaches 100%. Erlangs
1 # Trunks 50
October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 10
 Number of Trunks,
Capacity, and Utilization Efficiency

■ The graph at left illustrates

Capacity and Trunk Utilization the capacity in Erlangs of a
Erlang-B for P.02 Grade of Service given number of trunks, as
well as the achievable
90 45
40
utilization efficiency
80
70 35 ■ For accurate work, tables of
60 30 traffic data are available
50 25
40 20 • Capacity, Erlangs
30 15 • Blocking Probability
20 10 (GOS)
10 5
0 0 • Number of Trunks
0 10 20 30 40 50
Utilization Trunks Capacity,
■ Notice how capacity and
Efficiency Erlangs utilization behave for the
Percent
numbers of trunks in typical
cell sites

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 11

 Traffic Engineering & System Dimensioning

Using Erlang-B Tables to determine Number of Circuits Required

Probability
E of blocking
n 0.0001 0.002 0.02 0.2

1
2

7 2.935

Number of
Capacity
available
in Erlangs
circuits

300
A = f (E,n)

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 12

 Erlang-B Traffic Tables
Abbreviated - For P.02 Grade of Service Only

#Trunks Erlangs #Trunks

Erlangs #Trunks Erlangs#Trunks
Erlangs #Trunks Erlangs #Trunks
Erlangs #Trunks Erlangs #Trunks
Erlangs
1 0.0204 26 18.4 51 41.2 76 64.9 100 88 150 136.8 200 186.2 250 235.8
2 0.223 27 19.3 52 42.1 77 65.8 102 89.9 152 138.8 202 188.1 300 285.7
3 0.602 28 20.2 53 43.1 78 66.8 104 91.9 154 140.7 204 190.1 350 335.7
4 1.09 29 21 54 44 79 67.7 106 93.8 156 142.7 206 192.1 400 385.9
5 1.66 30 21.9 55 44.9 80 68.7 108 95.7 158 144.7 208 194.1 450 436.1
6 2.28 31 22.8 56 45.9 81 69.6 110 97.7 160 146.6 210 196.1 500 486.4
7 2.94 32 23.7 57 46.8 82 70.6 112 99.6 162 148.6 212 198.1 600 587.2
8 3.63 33 24.6 58 47.8 83 71.6 114 101.6 164 150.6 214 200 700 688.2
9 4.34 34 25.5 59 48.7 84 72.5 116 103.5 166 152.6 216 202 800 789.3
10 5.08 35 26.4 60 49.6 85 73.5 118 105.5 168 154.5 218 204 900 890.6
11 5.84 36 27.3 61 50.6 86 74.5 120 107.4 170 156.5 220 206 1000 999.1
12 6.61 37 28.3 62 51.5 87 75.4 122 109.4 172 158.5 222 208 1100 1093
13 7.4 38 29.2 63 52.5 88 76.4 124 111.3 174 160.4 224 210
14 8.2 39 30.1 64 53.4 89 77.3 126 113.3 176 162.4 226 212
15 9.01 40 31 65 54.4 90 78.3 128 115.2 178 164.4 228 213.9
16 9.83 41 31.9 66 55.3 91 79.3 130 117.2 180 166.4 230 215.9
17 10.7 42 32.8 67 56.3 92 80.2 132 119.1 182 168.3 232 217.9
18 11.5 43 33.8 68 57.2 93 81.2 134 121.1 184 170.3 234 219.9
19 12.3 44 34.7 69 58.2 94 82.2 136 123.1 186 172.4 236 221.9
20 13.2 45 35.6 70 59.1 95 83.1 138 125 188 174.3 238 223.9
21 14 46 36.5 71 60.1 96 84.1 140 127 190 176.3 240 225.9
22 14.9 47 37.5 72 61 97 85.1 142 128.9 192 178.2 242 227.9
23 15.8 48 38.4 73 62 98 86 144 130.9 194 180.2 244 229.9
24 16.6 49 39.3 74 62.9 99 87 146 132.9 196 182.2 246 231.8
25 17.5 50 40.3 75 63.9 100 88 148 134.8 198 184.2 248 233.8

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 13

 The Equation behind the Erlang-B Table

The Erlang-B formula is fairly simple to implement on

hand-held programmable calculators, in spreadsheets,
or popular programming languages.

Offered Traffic
lost due to
An max # of
blocking
trunks
n!
Pn(A) =
n
1 + A + ... + A
1! n! Number
of
Trunks average
Pn(A) = Blocking Rate (%) # of busy
with n trunks Offered channels
Traffic,
as function of traffic A
A
A = Traffic (Erlangs)
n = Number of Trunks time

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 14

 Wireless Traffic Variation with Time:
A Cellular Example
■ Peak traffic on cellular systems
is usually daytime business-
Typical Traffic Distribution related traffic; on PCS systems,
on a Cellular System evening traffic becomes much
100% more important and may actually
90% SUN contain the system busy hour
80%
MON ■ Evening taper is more gradual
70%
TUE
than morning rise
60%
50% WED
■ Wireless systems for PCS and
40%
LEC-displacement have peaks
THU
30% of residential traffic during early
20%
FRI evening hours, like wireline
10% SAT systems
0% ■ Friday is the busiest day,
Hour
followed by other weekdays in
backwards order, then Saturday,
then Sunday
■ There are seasonal and
Actual traffic from a cellular system in the annual variations, as well as
mid-south USA in summer 1992. This long term growth trends
system had 45 cells and served an area
of approximately 1,000,000 population.
October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 15
 Busy-Hour
■ In telephony, it is customary to collect and analyze traffic in hourly
blocks, and to track trends over months, quarters, and years
• When making decisions about number of trunks required, we
plan the trunks needed to support the busiest hour of a normal
day
• Special events (disasters, one-of-a-kind traffic tie-ups, etc.)
are not considered in the analysis (unless a marketing-
sponsored event)
■ Which Hour should be used as the Busy-Hour?
• Some planners choose one specific hour and use it every day
• Some planners choose the busiest hour of each individual day
(“floating busy hour”)
• Most common preference is to use “floating (bouncing)” busy
hour determined individually for the total system and for each
cell, but to exclude special events and disasters
• In the example just presented, 4 PM was the busy hour every
day
October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 16
 Where is the Traffic?

■ Wireline telephone systems have a big

advantage in traffic planning.
• They know the addresses where
their customers generate the traffic!
■ Wireless systems have to guess where
the customers will be next
• on existing systems, use
measured traffic data by sector and
cell
Existing System
Traffic In Erlangs – analyze past trends
8

2
5
7
7
11
10
– compare subscriber forecast
6 11
5 16 19
7
8 7
16 7 – trend into future, find overloads
6
3 9
9 • for new systems or new cells,
we must use all available clues

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 17

 Traffic Clues

27 mE/Sub in BH ■ Subscriber Profiles:

103,550 Subscribers • Busy Hour Usage, Call Attempts, etc.
1,239,171 Market Population
■ Market Penetration:
adding 4,350 subs/month
• # Subscribers/Market Population
Population Density • use Sales forecasts, usage forecasts
■ Population Density
• Geographic Distribution
■ Construction Activity
new
Shopping Center ■ Vehicular Traffic Data
• Vehicle counts on roads
Vehicular Traffic • Calculations of density on major
Land Use
920
5110 roadways from knowledge of vehicle
4215
Databases 22,100 movement, spacing, market
3620
1230
6620 penetration
■ Land Use Database: Area Profiles
■ Aerial Photographs: Count Vehicles!

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 18

 Traffic Density Along Roadways
Vehicles per mile ■ Number of lanes and speed are the main
variable determining number of vehicles on
Vehicle Vehicle Vehicles major highways
Speed, Spacing, per mile,
MPH feet per lane • Typical headway ~1.5 seconds
0 20 264 • Table and figure show capacity of 1
10 42 126
lane
20 64 83
30 86 61 ■ When traffic stops, users generally increase
45 119 44 calling activity
60 152 35 ■ Multiply number of vehicles by percentage
Vehicle spacing 20 ft. @stop penetration of population to estimate number
Running Headway 1.5 seconds of subscriber vehicles

Vehicle Spacing At Common Roadway Speeds

0 100 200 300 400 500 600 700 800 feet

0 MPH
10 MPH
20 MPH
30 MPH
40 MPH
50 MPH

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 19

 Methodical Estimation of Required Trunks

Modern propagation prediction

tools allow experimentation and
Traffic
Density
estimation of traffic levels
3.5%
■ Estimate total overall traffic from
subscriber forecasts
■ Form traffic density outlines
27mE from market knowledge,
Land Use forecasts
■ Overlay traffic density on land
use data; weight by land use
■ Accumulate intercepted traffic
into serving cells,
• obtain Erlangs per cell &
Cell Grid sector
■ From tables, determine number
of trunks required per cell/sector
■ Modern software tools automate
major parts of this process
October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 20
 Profile of Typical Cellular Usage

Offered Traffic, mE per subscriber in busy hour 25 mE

Number of call attempts per subscriber in busy hour 1.667
Average Call Duration 150 sec. (41.7 mE)
Mobile originated calls
proportion of total calls on system 87 %
successful calls 70 %
Calls not answered 15 %
calls to a busy line 15 %

Mobile terminated calls

proportion of total calls on system 13 %
successful calls 15 %
Calls not answered 10 %
paging requests not answered 75 %
Number of handoffs per call 0.87
Registration attempts per subscriber during busy hour 2

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 21

 Determining Number of Trunks
required for a new Growth Cell
When new growth cells are added, they absorb
some of the traffic formerly carried by
surrounding cells
■ Two approaches to estimating traffic on the
new cell and on its older neighbors:
• if blocking was not too severe, you
can estimate redistributed traffic in the
area based on the new division of
coverage
• if blocking is severe, (often the case),
users will stop trying to call in locations
where they’ve learned to expect
blocking. Users are programmable!!
– reapply basic traffic assumptions in
the area, like engineering new
system, for every nearby cell
– watch out! overall traffic in the area
may increase to fill the additional
capacity and the new cell itself may
block as soon as it goes in service

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 22

 Dimensioning
System Administrative Functions

System administrative functions also require traffic engineering input. While

these functions are not necessarily performed by the RF engineer, they
require RF awareness and understanding.
■ Paging
• The paging channel has a definite capacity which must not be
exceeded. When occupancy approaches this limit, the system must
be divided into zones, and zone paging implemented.
■ Autonomous Registration
• Autonomous registration involves numerous parameters and the
registration attempts must be monitored and controlled to avoid
overloading.
■ Locate Measurements
• Locate measurement requests can overrun the capacity of locate
receivers in specific cells. This must be monitored in busy systems
and controlled by careful attention to cell-pair definitions and trigger
thresholds. Course 1001, Handoff Seminar, provides more details.

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 23

 Trunking Efficiency
An Important Cellular Consideration

■ Busy cellular systems often use

sectorized cells
• A cell’s coverage area is divided
into several “sectors” using
directional antennas
45 channels – 3-sector (120-degrees)
– 6-sector (60-degrees)
• radio channels assigned per
sector
■ Capacity of a sectorized cell is less
than capacity of an omni cell with
same total number of channels
• 45 channels: 35.61
15 ch. Erlangs
15 ch. • 3 x 15 channels: 3 x 9.01 Erl.
15 ch. = 27.03 Erlangs
■ Why would anyone sectorize?
• Sectorization eases frequency
reuse more than it hurts capacity

October, 1997 RF Engineering 102 v1.0 (c)1997 Scott Baxter 6 - 24