Gladiator Startup 1.0

Download Report

Transcript Gladiator Startup 1.0

ECE 5221 Personal Communication Systems
Prepared by:
Dr. Ivica Kostanic
Lecture 17: Traffic planning
Spring 2011
Florida Institute of technologies
Outline
Traffic in communication networks
Circuit switched versus packet switched traffic
Queuing system
Elements of queuing system
Traffic in erlangs
Important note: Slides present summary of the results. Detailed
derivations are given in notes.
Florida Institute of technologies
Page 2
Traffic in communication networks
 Traffic - flow of information messages
through a communication network
 Generated as a result of
o phone conversations
o data exchange
o audio, video delivery
o signaling
 Communication networks are designed to
provide service to many users
 At any instant of time not all users are
active
o network resources are shared
o resource sharing may result in
temporary service unavailability
 Traffic planning allows sharing of
resources with minimum performance
degradation
Communication Network
Modern communication networks carry
mixture of voice and data traffic
Florida Institute of technologies
Page 3
Outline of a cellular network
BTS
BSC
MSC
MSC
BSC
BTS
BTS
BTS
BTS
BTS
BTS - Base Station
BSC - Base Station Controler
MSC - Mobile Switchning Center
 Cellular network consists of many
connected elements
 Analysis of the entire network is
complicated
o Common practice - analyze each
link individually
 Traffic dimensioning has two aspects
o Dimensioning the network elements to
have enough processing power
o Dimensioning the connecting lines to
have sufficient capacity
Traditionally, traffic bottleneck - Air interface
Florida Institute of technologies
Page 4
Circuit switched communication services
 First and second generation provides connection oriented services to the users
 A dedicated channel is allocated over the entire duration of the call
 In the case of voice communication this is “only” 50% wasteful
 This mode of communication is called “circuit-switching”
 Circuit switching is very inefficient for data communication (major driver of 3G
cellular systems)
 Circuit switching is abandoned in 4G
Interpretation of term circuit for various cellular technologies
Technology
Circuit resource
FDMA/TDMA
Pair of frequencies and associated time slot
FDMA/CDMA
Pair of frequencies + associated codes
Florida Institute of technologies
Page 5
Packet switched communication services
 Datagram packet switching
 Virtual path packet switching
o Every packet travels independently
o Virtual path (sequence of network
nodes) is established through the
network
o Implemented within IP based networks
o Implemented within ATM networks
o Transport layer has to assure the proper
order of the packets
2
4
Virtual Path
6
3
7
1
7
2
6
3
8
5
1
5
8
4
Virtual path switching
Datagram switching
Note: Modern packet data networks are using datagram switching
Florida Institute of technologies
Page 6
Types of traffic in cellular networks
 Cellular networks support
ITU vision for cellular services
o circuit switched (CS) voice
o dispatch voice (push to talk)
o circuit switched data
o packet data (PD)
 Communication resources may be
o Shared between CS and PS
o Separated resources may be set for
CS and PS
 First and second generation dominated with circuit switched voice
 Third generation and beyond dominated by data
Traffic planning in heterogeneous
cellular networks of the future takes
central stage
Florida Institute of technologies
Page 7
Description of queuing systems
 Queuing systems
Mean
Arrival
Rate
l
Number of
Users in the
Queue
Nq
o Mathematical abstraction
S1
S2
 Elements of a queuing
system
Interarrival
Time
t
o Used to develop the traffic
analysis and planning
methodology
Queuing
Time
q
Sc
o source population
o queue
Source
Population
Generated
Traffic
Queue
Servers
Outline of a queuing system
o servers
o distributions of interarrival
times, service times,
queuing discipline, etc.
 Queuing system – cell site
 Servers – channel resources – trunks
 Population – users connecting to cellular network
Florida Institute of technologies
Page 8
Source population
 Consists of all users that are eligible for service
 The most important property - size
o infinite population - arrival rate does not
depend on the number of users in the system
o finite population - arrival rate depends on the
number of users in the system
o if the population is large relative to the number
of servers we routinely assume that its is infinite
 In cellular systems population are all eligible
users within the coverage area of the cell
 It is assumed that the number of eligible users
is much greater than the number of the users
using the system at any given moment
 Over a course of day, the size of population
changes
Example of a call stats
benchmarking map
 Traditionally cellular systems are dimensioned
for a good performance during the busiest hour
Florida Institute of technologies
Page 9
Arrival rate and interarrival times
 Arrival rate - number of service
requests per unit time
Example:
 The ability of the queuing system to
provide effective service depends on
distribution of arrival rates
The average number of call arrivals in two
figures is the same: 20 arrivals per minute.
The traffic pattern in second figure requires
more resources to accommodate for higher
demand peaks.
 Standard way of specifying arrival rate
is through probability density function
of interarrival times
50
40
40
number of call attempts
number of call attempts
50
30
30
20
20
10
0
0
10
10
20
30
time [min]
40
50
60
0
0
10
20
Florida Institute of technologies
30
time [min]
40
50
60
Page 10
Service time (call holding time-CHT)
Example: Duration of CHT at a cell
 Service time-period of
time that the resource
is allocated to
individual user
 Most commonly, CHT
is exponentially
distributed
relative frequency of occurance
 Usually specified
through its distribution
Histogram of call holding time (CHT), mean = 91.6s, std = 95.08sec, 498 measurements
0.015
Exponential PDF, mean = 92 sec
0.01
0.005
Exponential distribution
pdf exp  x  
1
 x
exp   , x  0
T
 T
0
0
100
200
300
400
500
600
call duration [sec]
T – average call holding time
Note: Exponential distribution is a good model for demand generated by
humans (voice, SMS, email,..)
Florida Institute of technologies
Page 11
Average resource occupancy - traffic in erlangs
Definition
 Erlang - unit for measuring of
traffic intensity
a
 Defined as a fraction of time
that the resource is occupied
 Occupancy does not have to
continuous
 Specified relative to some
averaging time
 Maximum traffic carried by a
single resource - 1 erlang
t
T
[erlang]
t
Resource occupancy time
T
Averaging time
Example
t1Average traffic
 Total traffic carried by service
facility cannot exceed number
of servers
t2
t3
T
A
t 1 + t 2 + t 3 1 . 5 + 2 + 1 4 .5


 0.5635 E
T
8
8
Florida Institute of technologies
Page 12
Alternative interpretation of erlang traffic
 Traffic in erlangs = average number of
simultaneously occupied servers
Traffic in erlangs for multi-server system
 Can be measured easily
C
tc
t
t1
t2
A  1 + 2 + + C   n n
T
T
T n 1 T
C
T
10
Sum of times during exactly n out of
C servers are held simultaneously
Number of servers
Averaging time
number of occupied channels
tn
o regular poling of service facility and logging
the number of occupied resources
8
Average traffic
6
4
2
0
0
10
20
30
time [min]
40
50
60
Example of traffic measurements.
Averaging time is 60 min.
Poling time is 1 min.
Florida Institute of technologies
Page 13
Offered, carried and lost traffic
Relation between offered, carried and lost traffic
 Lost traffic - traffic that could not
be served due to finite resources
 Served traffic - difference
between offered and lost traffic
 Attempt to serve all offered traffic
results in allocation of large
number of resources
18
Number of channels
Lost traffic
16
number of occupied channels
 Offered traffic - traffic that would
be served if the number of
resources is unlimited
Offered traffic
14
12
10
8
6
4
2
0
10
20
30
time [min]
40
50
60
Note : Communication systems are frequently designed to operate with
a certain percentage of lost traffic
Florida Institute of technologies
Page 14