Telecoms Network Design
T-110.300
Timo Kiravuo, partially from slides by
Olli Knuuttila & Hannu H. Kari
Design Issues
• Telecoms networks are expensive
– Installation in physically locations
– Sometimes access rights and physical access is difficult
– High availability hardware
– Time to market is an important factor
– Heavy investment prior to returns
– Long payback time
• Income is based on usage
– If the available resources can not fulfill a demand, income is
lost
– Lacking customer satisfaction increases churn (loss of
customers)
• Wrong investments eat profit margins and can destroy the
company
Planning and Dimensioning
• Inputs
– User needs
– Current statistics
– Future estimates
– Competition
– Technology, standards, products
– Visioning
• Internals
– Capacity needs
– Network element requirements
– Modeling network / user behavior
• Outputs
– Strategic planning, company policy
– Installation/purchase schedule
– Configuration setting
– Prioritizing, pricing rules
Dimensioning Examples (GSM)
• Capacity
– Number of users supported in the network
– Number of lines between BSC and MSC
– Number of carriers per cell
– Number of modem lines for GSM data
– Disk space and CPU capacity for voice mail
• Location
– Where to install new equipment
– Central or distributed service topology
– Location of the base stations
– Location of the modem pool
Load Control
• In the circuit switched world service is black & white
– Call is established or blocked
– Packet networking is different
• How to reduce load in the network
– Blocking -> unhappy customers
– Pricing -> happy customers
• In mobile networks increasing load decreases quality
– In TDMA other mobiles are monitoring same frequency and
time slot
– In CDMA other mobiles are monitoring all other mobiles in the
same cell
– Fixed line telephone networks are more simple
Increasing Capacity (GSM)
• Radio capacity
– More base stations, smaller cell radius
– Costly, usually possible
– More frequencies / channels
– Depends on regulation or base station hardware
– Better radio frequency utilization through different coding
– Depends usually on handset hardware and standards
• More efficient utilization of resources
– Circuit vs. packet switched services
– Dedicated vs. multiplexed channels
• Parallel servers (pizza boxes)
– GGSN, SGSN, MSC, HLR
– requires algorithms that parallelize
Busy Hour
• The sliding 60 minute period, during which occurs the
maximum total traffic load in a given 24 h period
• Examples:
– Morning: 9-11
– Afternoon 13-17
– Evening: 19-21
• Operator network design usually focuses on the Busy Hour
– The load of off-peak hours is not significant for network design
• If traffic can somehow be moved from peak hours to off-
peak hours, operators can save on network costs
– Pricing
Traffic Measurements
• Traffic is measured in Erlangs
– Strict definition is related to one voice path
• In network design Erlang is usually used to
measure traffic volume in one hour
– If a set of subscribers makes 30 calls in one hour and
the average call is 5 minutes, that is 2.5 Erlangs
• Named after A. K. Erlang, who did the basic
research on this in the early 1900's
• Applies to services as well as trunk lines
Customer Profiling
• E.g. if an average residential customer creates 25 mErl
during the busy hour, 50 000 subscribes would generate
1250 calls (Erlangs) on the average
• Profiles depend on culture and area, change with time, are
affected by mobility etc.
• Traffic patterns can be studied from the logs of carried traffic
– Notice that blocked calls are not shown
• Sometimes traffic is difficult to predict
– Major accidents
– TV-competitions
– National holidays
Blocking
• Blocking means how many uncompleted calls do
we accept
– Usually during the Busy Hour
• Measured usually as a percentage
– 1% blocking during the Busy Hour is often acceptable
– ITU-T E.543
– Means that 1 out of 100 calls can not be completed,
because the network is full and no free path can be
found
– 1% is the design target, most of the time
performance is much more better
• Blocking can be caused by lack of trunk lines,
congestion in signaling, unavailable services etc.
The Erlang B Formula
• The Erlang B Formula calculates the probability of a call
being blocked, based on the number of links (trunks) or
other available services and offered load
• The call arrivals are assumed to be random (Poisson
distribution)
c
a / c!
Pb ? B (c, a) ? c
• c = trunks, operators etc.
? k? 0
k
( a / k! )
• a = offered load
• Pb = Probability of call being blocked
The Erlang B Formula in Practice
• Increasing the amount of trunks decreases
the blocking probability in non-linear fashion
– This table shows the Erlang-value 0.5% 1% 2%
as a function of trunk lines and the 1 0.005 0.010 0.020
blocking factor
2 0.105 0.153 0.223
– To carry 1 Erl at 1% blocking,
3 0.349 0.455 0.602
requires 5 lines
4 0.701 0.869 1.092
– 8 lines can carry 3 Erls
5 1.132 1.361 1.657
– Accepting greater blocking
likelihood allows some more traffic 6 1.622 1.909 2.276
– Note that the amount of users is 7 2.157 2.501 2.935
not important, traffic generated is 8 2.730 3.128 3.627
Practical Network Dimensioning
• Amount of traffic can be estimated from
– Past experience
– Known demographics
– City planning
– etc.
• In fixed line networks the number of trunks is one major
issue
– These come usually in PDH multiplexes or SDH containers
– 30 trunks can carry 20 Erl at 1% blocking
– 120 trunks can carry 103 Erl at 1% blocking
• There is also the switching equipment and services
Network Throughput
• The issue of interest to users is the quality of the total
service
– The end to end traffic
– Calls cross the telephone network over many trunks and
switches, any of which may be blocked
• Some network patterns are easy to analyze
– A link, series of links, a tree
• The actual mesh network is hard to analyze
• One method is to analyze all the major paths
– E.g. all trunks have a blocking factor of 1%, if a call crosses
three trunks, the total factor is 1-(1-0.01)3 = 3%
– ITU-T has standardized both the reference network and QoS
requirements
More Network Design Issues
• Signaling network capacity should also be
calculated
• And switching capabilities
System capabilities
Traffic volume Quality of Service
Erlangs B < 1%
Network Topology
• Fixed line networks are usually meshes built of
rings
– Redundancy and alternative routes
– Alternatives can also carry excess traffic
– Land rights and equipment center permissions are
important
• Cellular networks provide a completely different
challenge
– Urban areas
– Dense population, buildings
– Rural areas
– Mobility
– Dense lines across rural areas
– Unpredictable traffic
Cellular Network Topology
• Cellular network design is focused on the
– Transceiver antenna locations and directions
– Radio frequencies and channels
– Transmission power
• Trunks from the base station to backbone are also
an issue, but more manageable one
– Dual band GSM requires roughly 2-4 Mbps
– Microwave links
• Cells can be macro, micro or indoor cells
– Smaller cells enable better re-use of frequencies but
cost more
Regulation
• Government assistance requirements
• Emergency services
• Consumer protection
– QoS requirements
– Monopoly regulation
– Mandatory service provisioning
– Customers
– Competition
• Resources
– Frequencies
– Right of way
• No control over terminal equipment or service usage
