Traffic Engineering

1 Efficiency % 80% 41 Capacity, Erlangs 50 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

Traffic by Hour

Hour

Objectives

• Identify the role and functions of traffic engineering in a wireless system • Understand the basic units and concepts of traffic engineering • Understand basic principles of system dimensioning • Develop operational familiarity with traffic tables • Understand and apply the concept of trunking efficiency to wireless systems • Examine current methods of wireless traffic forecasting and analysis

Outline

• • • • • • Traffic Engineering and System Dimensioning Objectives Basics of Traffic Engineering – trunk concept – units of traffic measurement – offered traffic and call duration – blocking probability and grade of service – capacity and utilization efficiency as a function of number of trunks Traffic Tables and Formulas Variation of Traffic with Time – Real-System Example and Busy-hour determination – Typical Traffic Profile of Cellular System Traffic Estimation and Cell Trunk Dimensioning – Geographic Distribution of Traffic and its estimation • for new systems, for growth cells Exercise

$

Traffic Engineering Objectives

3 7 1 9 9 3 1 6 8 2 7 3 4 9 2 5 1 8 2 4 7 8 6 11 6 10

• • • Traffic engineering is the intelligent art of having adequate capacity, but not spending too much to get it Traffic engineering is applied during every stage in the development and operation of a cellular system In Initial Design: – How many cells are needed?

– What about switching resources?

– What is the optimal way to backhaul?

Ongoing during Operation: – What BTS resources, and when?

– When are new BTSs needed?

– Anticipate resource requirements to allow budgeting and installation

Walking a Fine Line

BTS BTS BTS BTS BTS BTS BTS BTS BTS BTS BTS BTS BTS BTS BTS

• • 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

Underdimensioning

– (

blocking

– poor technical performance

interference

) – capacity for billable revenue is low – revenue is low due to poor quality – users unhappy, cancel service

Basics of Traffic Engineering Terminology & Concept of Trunks

• • • Traffic engineering in telephony is focused on the

voice paths users occupy

. They are called by various names: –

trunks

– –

circuits voice paths

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 In a CDMA system, the air interface is soft- squeezed. But there are other hard resources to be dimensioned: – Vocoders in the BSC – Channel elements in the BTS

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 develop the concepts of trunk dimensioning and publish the information for the benefit of 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 normally one hour long -- I.e.,

one hour of talk

.

• • • Other units have become popular among various users:

CCS MOU

(Hundred-Call-Seconds) (Minutes Of Use) It is easy to convert between traffic units if the need arises:

1 Erlang = 60 MOU = 36 CCS

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 cellular system: – not enough channel elements, the BTS is full – not enough vocoders in the BSC – not enough paths through the switching complex – not enough trunks from switch to PSTN Blocking probability is usually expressed as a percentage :

P.02

is 2% probability, etc.

– Blocking probability sometimes is called

g

rade

O

f

S

ervice Most blocking in cellular systems occurs at the BTS level.

P.02 is a common goal

MTX P.005

Design Blocking Probabilities PSTN Office BSC BTS BTS P.02

BTS P.001

P.005

Offered and Carried Traffic

BTS PSTN or other Wireless user Carried Traffic MTX BTS BSC BTS BTS Offered Traffic BTS BTS Blocked Traffic

• • •

Offered Traffic

originate.

is what users attempt to

Carried Traffic

is the traffic actually successfully handled by the system

Blocked traffic

is the traffic that could not be handled – since blocked call attempts never materialize, blocked traffic can only be estimated based on number of blocked attempts and average duration of successful calls

T

O

= CA x CD

T O

= offered traffic

(any desired units)

CA

= total call attempts

CD =

average successful call duration (T O and CD must be in same units)

Traffic Engineering and Queuing Theory

Ticket Counter Analogy

Servers Queue User Population

Queues we face in Everyday Life 1) for telephone calls, 2) at the bank 3) at the gas station 4) at the airline counter

• • • • • • Traffic Engineering is an application of a science called

queuing theory

Queuing theory relates user arrival statistics, number of servers, and various queue or waiting strategies, with the probability of a user receiving service meeting specified criteria If waiting is not allowed, and a blocked call simply goes away,

Erlang-B

formula applies

(popular in wireless)

If unlimited waiting is allowed before service, the

Erlang-C

formula applies If a wait is allowed but is limited in time, blocked calls held, Binomial & Poisson formulae apply Fast, short transactions with no wait allowed:

Engset

formula applies

Number of Trunks vs. Utilization Efficiency

• • • • Imagine a BTS with

just one voice channel

. It can carry an erlang. But 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%!

Erlang-B P.02 GOS

Adding just one trunk relieves things greatly. Now we can use

Trks

trunk 1 heavily, with trunk 2 handling the overflow.

Efficiency rises to 11%

1 2 Erl 0.02

0.22

Eff% 2% 11%

•

The Principle of Trunking Efficiency For a given grade of service, trunk efficiency increases as the the pool grows larger.

– For trunk groups of several hundred, approaches 100%.

number of trunks in

41 1 utilization Capacity, Erlangs # Trunks 80% 50

Number of Trunks, Capacity, and Utilization Efficiency

Capacity and Trunk Utilization Erlang-B for P.02 Grade of Service

90 80 70 60 50 40 30 20 10 0 0 Utilization Efficiency Percent 10 20 Trunks 30 40 45 40 35 30 25 20 15 10 5 0 50 Capacity, Erlangs • • • The graph at left shows the capacity in erlangs for a given number of trunks, as well as the utilization efficiency For accurate work, tables of traffic data are available – Capacity, Erlangs – Blocking Probability (GOS) – Number of Trunks Notice how capacity and utilization behave for the numbers of trunks in typical cell sites

Traffic Engineering & System Dimensioning

Using Erlang-B Tables to determine Number of Circuits Required

Probability of blocking n E 1 2 0.0001 0.002

0.02

0.2

Number of available circuits 7 300 2.935

Capacity in Erlangs

A = f (E,n)

Erlang-B Traffic Tables

Abbreviated - For P.02 Grade of Service Only

#TrunksErlangs 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 0.0204

0.223

0.602

1.09

1.66

2.28

2.94

3.63

4.34

5.08

5.84

6.61

7.4

8.2

9.01

9.83

10.7

11.5

12.3

13.2

14 14.9

15.8

16.6

17.5

18.4

19.3

20.2

21 21.9

22.8

23.7

24.6

25.5

26.4

27.3

28.3

29.2

30.1

31 31.9

32.8

33.8

34.7

35.6

36.5

37.5

38.4

39.3

40.3

#TrunksErlangs 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 41.2

42.1

43.1

44 44.9

45.9

46.8

47.8

48.7

49.6

50.6

51.5

52.5

53.4

54.4

55.3

56.3

57.2

58.2

59.1

60.1

61 62 62.9

63.9

#Trunks Erlangs 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 64.9

65.8

66.8

67.7

68.7

69.6

70.6

71.6

72.5

73.5

74.5

75.4

76.4

77.3

78.3

79.3

80.2

81.2

82.2

83.1

84.1

85.1

86 87 88 #TrunksErlangs #TrunksErlangs 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 88 89.9

91.9

93.8

95.7

97.7

99.6

101.6

103.5

105.5

107.4

109.4

111.3

113.3

115.2

117.2

119.1

121.1

123.1

125 127 128.9

130.9

132.9

134.8

100 102 104 106 108 110 112 114 116 118 120 122 124 126 128 130 132 134 136 138 140 142 144 146 148 #TrunksErlangs 150 152 154 156 158 160 162 164 166 168 170 172 174 176 178 180 182 184 186 188 190 192 194 196 198 136.8

138.8

140.7

142.7

144.7

146.6

148.6

150.6

152.6

154.5

156.5

158.5

160.4

162.4

164.4

166.4

168.3

170.3

172.4

174.3

176.3

178.2

180.2

182.2

184.2

#TrunksErlangs 200 202 204 206 208 210 212 214 216 218 220 222 224 226 228 230 232 234 236 238 240 242 244 246 248 186.2

188.1

190.1

192.1

194.1

196.1

198.1

200 202 204 206 208 210 212 213.9

215.9

217.9

219.9

221.9

223.9

225.9

227.9

229.9

231.8

233.8

#TrunksErlangs 250 300 350 400 450 500 600 700 800 900 1000 1100 235.8

285.7

335.7

385.9

436.1

486.4

587.2

688.2

789.3

890.6

999.1

1093

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.(factorial)

max # of trunks Offered Traffic lost due to blocking A n n!

P n (A) = 1 + + ... + 1!

A n n!

P n (A) = Blocking Rate (%)

with n trunks as function of traffic A

A = Traffic (Erlangs) n = Number of Trunks Number of Trunks Offered Traffic, A average # of busy channels time

Wireless Traffic Variation with Time

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Typical Traffic Distribution on a Cellular System Hour SUN MON TUE WED THU FRI SAT Actual traffic measured on a system in the mid-south USA in summer 1992. This system had 45 cells and served an area of approximately 1,000,000 population.

• Peak traffic on earlier cellular systems was usually daytime business-related traffic – Evening taper is more gradual than morning rise – Friday is the busiest day, followed by other weekdays in backwards order, then • Saturday, then Sunday • Wireless systems for PCS will have peaks of residential traffic during early evening hours, like wireline systems

There are seasonal and annual variations, as well as long term growth trends

The 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 –

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

(or

bouncing

) busy hour determined individually for each cell and for the total system, but excluding one-of-a-kind events and disasters – In example chart just presented, 4 PM was the busy hour every day

High-Level Traffic Forecasting and Business Planning

Year

Population Penetration # Subscribers BH Erlangs/Sub BH Total Erlangs # of Cell Sites Avg. Erl/Cell Avg. Chan./Cell Total Voice Chans.

0/Start

1,000,000 0.1% 1,000 .100

100 10 10 17 170

1

1,000,000 2.5% 25,000 .05

1,250 25 50 61 1,525

2

1,000,000 5.0% 50,000 .04

2,000 35 57 68 2,380

3

1,000,000 7.0% 70,000 .03

2,100 40 52.5

64 2,560

4

1,000,000 9.0% 90,000 .028

2,520 45 56 67 3,015

5

1,000,000 12% 120,000 .025

3,000 50 60 71 3,550 • Every system deserves a business plan based on marketing and traffic assumptions. – The plan is used for forecasting equipment and capital needs.

– The number of cells is driven both by coverage needs and the requirement to carry anticipated traffic without blocking.

– Total is based on anticipated per-subscriber average usage & number of subs – The distribution of voice channels among cells is determined later.

Where is the Traffic?

Existing System Traffic In Erlangs 5 2 7 8 6 8 5 7 11 7 16 7 6 16 3 11 19 9 9 10 7

• • 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 • analyze past trends • compare subscriber forecast • trend into future, find overloads – for

new

systems or new cell, we must use all available clues

Traffic Clues

27 mE/Sub in BH 103,550 Subscribers 1,239,171 Market Population adding 4,350 subs/month Population Density Land Use Databases new Shopping Center Vehicular Traffic 920 5110 4215 22,100 3620 1230 6620

• • • • • • • Subscriber Profiles: – Busy Hour Usage, Call Attempts, etc.

Market Penetration: – # Subscribers/Market Population – use Sales forecasts, usage forecasts Population Density – Geographic Distribution Construction Activity Vehicular Traffic Data – Vehicle counts on roads – Calculations of density on major roadways from knowledge of vehicle movement, spacing, market penetration Land Use Database: Area Profiles Aerial Photographs: Count Vehicles!

Traffic Density along roadways

Vehicles per Mile

Vehicle Speed, MPH

0

Vehicle Spacing, feet

20

Vehicles per mile, per lane

264 10 20 42 64 126 83 30 45 60 86 119 152 61 44 35

Vehicle spacing 20 ft. @stop Running Headway 1.5 seconds

• • • Speed is the main variable determining number of vehicles on major highways – typical headway ~1.5 seconds – table and figure show capacity of 1 lane When traffic stops, users generally increase calling activity Multiply number of vehicles by percentage penetration of population to estimate number of subscriber vehicles

0 VEHICLE SPACING AT COMMON ROADWAY SPEEDS 100 200 300 400 500 600 700 800 feet 0 MPH 10 MPH 20 MPH 30 MPH 40 MPH 50 MPH

Systematic Estimation of Required Trunks

Traffic Density 3.5% Land Use 27mE Cell Grid

Modern propagation prediction tools allow experimentation and estimation of traffic levels • Estimate total overall traffic from subscriber forecasts • Form traffic density outlines from market knowledge, forecasts • Overlay traffic density on land use data; weight by land use • Accumulate intercepted traffic into serving cells, – obtain erlangs per cell & sector • From tables, determine number of trunks required per cell/sector • Modern software tools automate major parts of this process

Example Wireless Usage Profile

Offered Traffic, mE per subscriber in busy hour Number of call attempts per subscriber in busy hour Average Call Duration Mobile originated calls proportion of total calls on system successful calls Calls not answered calls to a busy line Mobile terminated calls proportion of total calls on system successful calls Calls not answered paging requests not answered Percentage of Time in Soft Handoff Registration attempts per subscriber during busy hour 25 mE 1.667

150 sec. (41.7 mE) 87 % 70 % 15 % 15 % 13 % 15 % 10 % 75 % 35% 2

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 estimate traffic on the new cell and on its older neighbors: – –

if blocking was not too severe

estimate redistributed traffic in the area based on the new division of coverage , you can

if blocking was severe

every nearby cell , (often the case), users may have quit trying to call in locations where they expected blocking • reapply basic traffic assumptions in the area, like engineering new system, for • 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

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 channels have a definite total capacity which must not be exceeded. When occupancy approaches this limit, the system must be divided into smaller zones, and registration parameters adjusted – Autonomous registration involves numerous parameters and the registration attempts must be monitored and controlled to avoid overloading.

Access Attempts

– Access attempts must be monitored and the number of enabled access channels set appropriately. On busy systems, probing sequence parameters should be closely observed and optimized

Trunking Efficiency An Important CDMA Implication

CONVENTIONAL SECTORIZATION 1 1/3 1/3 1/3 CDMA SECTORIZATION 1 1 1 1

• • AMPS/TDMA/GSM sectorization distributes available channels among sectors – this results in a net

decrease

cell total cap. = 27.01 Erl in capacity, although it gives better flexibility for managing interference – Example: 45 ch. omni = 35.6 Erl 3=sector: 15 ch. = 9.01 Erl, sector In CDMA, each additional sector is an additional independent signal – Each additional sector has almost as much capacity as the original omni configuration!

– Inter-sector boundary interference places a practical limit somewhat above 6 sectors

Digital Transmission Hierarchy

• • • • The digital signal hierarchy is the foundation of the PSTN:

DS-0

– a two-way 64 kilobit/second digital circuit that carries a single conversation

DS-1 (sometimes called a E

– you can lease a E 1

1 circuit)

– a 2.048 megabit/second combination of 30 DS-0s: it carries 30circuits from a LEC or an IXC, or you can build microwave to haul DS-1s

DS-3

– a 34 megabit/second combination of 16 DS-1s: 16X30= 480 circuits – you can lease a DS-3 from a LEC or an IXC or you can build microwave to haul DS-3s

30 DS-0s Optical Formats

: OC1, OC3, OC24, OC48 – OC-1 =55 mb/s

16 DS-1s 1 DS-0 30 DS-1 T-1 480 DS-3 30 DS-0s 16 DS-1s

Traffic Engineering Exercise

Busy-Hour Contribution 0.2 Erlangs 1.0 Erlangs 3.0 Erlangs

Traffic Engineering Exercise

Busy-Hour Contribution 0.2 Erlangs 1.0 Erlangs 3.0 Erlangs

• • • • • • Imagine you have been assigned to plan this new cellular system. You have already predicted traffic densities, and set up the cell grid to meet basic requirements and fit with adjoining systems.

The matrix of squares ( represents predicted busy hour offered traffic.

The hexagonal cell grid represents the coverage areas of planned cells

Q:

For each cell, determine the traffic intercepted, and the number of channels needed to give a P.02 Grade of Service

Q:

In 12 months, traffic is expected to increase 40%. Then, what will be the channel requirements for each cell?

Refer to the preceding page for a readable traffic density diagram. Following pages provide forms to help speed and organize your work.

Form 1 for Traffic Engineering Exercise

• Suggested Procedure: – Refer to the enlarged traffic density diagram on an earlier page.

– For each cell, tally its intercepted bins of each type.

– Convert bin counts into Erlangs.

– Determine channels required from the P.02 Erlang B table

Traffic Bin Counts Traffic, Erlangs Channels Required 0.2 1.0 3.0 Erlangs

(see next page to continue for traffic growth part)

Form 2 for Traffic Engineering Exercise

• • At the end of 12 months, traffic will have increased 40%.

Suggested procedure: – Multiply your originally-determined Erlang figures by 1.4, then – Determine channels required from the P.02 Erlang B table

Original Erlangs New Erlangs Channels Required

Traffic Engineering Solution, Part 1

• One bin-counter answers are shown below.

• Your bin count may differ slightly, depending on how you resolve close cases, but your channel answers should be approximately the same as shown.

7 15 7 19 3 17 5 -

Traffic Bin Counts

6 6 11 2 13 12 2 16 2 2 3 6 4 5 17 9 5 3 11 2 12 9 5 14 13 1 12 11 3 19 4 6 2 5 8 7 1 13 3 13 3 -

Traffic, Erlangs 1.4

4.2

3.2

4.2

10 6 12.4

11.6

6.8

8.4

20.6

21 19 14 26.4

35.2

18.8

22.4

5.6

5.6

7.8

1.0

Channels Required 5 9 8 9 17 13 15 12 20 14 19 29 45 29 27 21 35 27 31 11 11 4 0.2 1.0 3.0 Erlangs

(see next page to continue for traffic growth part)

Traffic Engineering Solution, Part 2

• An exercise like this increases appreciation of software tools!

• Traffic increased 40%; the required channel increases ranged from 25% in the smallest cells to 34% in the biggest cells.

• Food for thought: – How many CDMA carriers are needed for each sector?

– Can you do a PN offset plan for this system?

Original Erlangs New Erlangs Channels Required 10 6.8

8.4

1.4

4.2

3.2

4.2

12.4

20.6

21 19 14 26.4

35.2

18.8

22.4

6 7.8

11.6

5.6

5.6

1.0

1.96

5.88

4.48

5.88

14 8.4

17.36

16.24

9.52

11.76

28.84

29.4 36.96

19.6

26.6

7.84

49.28

31.36

26.32

7.84

10.92

1.4

21 16 19 6 12 38 60 15 12 25 39 36 28 47 35 41 18 10 24 14 14 5