1. Florida Institute of technologies ECE 5221 Personal Communication Systems Prepared by: Dr. Ivica Kostanic Lecture 17: Traffic planning Spring 2011

2. Florida Institute of technologies Page 2  Traffic in communication networks  Circuit switched versus packet switched traffic  Queuing system  Elements of queuing system  Traffic in erlangs Outline Important note: Slides present summary of the results. Detailed derivations are given in notes.

3. Florida Institute of technologies Page 3 Traffic in communication networks  Traffic - flow of information messages through a communication network  Generated as a result of ophone conversations odata exchange oaudio, video delivery osignaling  Communication networks are designed to provide service to many users  At any instant of time not all users are active onetwork resources are shared oresource sharing may result in temporary service unavailability  Traffic planning allows sharing of resources with minimum performance degradation Modern communication networks carry mixture of voice and data traffic

4. Florida Institute of technologies Page 4 Outline of a cellular network  Cellular network consists of many connected elements  Analysis of the entire network is complicated oCommon practice - analyze each link individually  Traffic dimensioning has two aspects oDimensioning the network elements to have enough processing power oDimensioning the connecting lines to have sufficient capacity Traditionally, traffic bottleneck - Air interface

5. Florida Institute of technologies Page 5 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 TechnologyCircuit resource FDMA/TDMAPair of frequencies and associated time slot FDMA/CDMAPair of frequencies + associated codes

6. Florida Institute of technologies Page 6 Packet switched communication services  Virtual path packet switching oVirtual path (sequence of network nodes) is established through the network oImplemented within ATM networks  Datagram packet switching oEvery packet travels independently oImplemented within IP based networks oTransport layer has to assure the proper order of the packets Virtual path switchingDatagram switching Note: Modern packet data networks are using datagram switching

7. Florida Institute of technologies Page 7 Types of traffic in cellular networks  Cellular networks support ocircuit switched (CS) voice odispatch voice (push to talk) ocircuit switched data opacket data (PD)  Communication resources may be oShared between CS and PS oSeparated resources may be set for CS and PS  First and second generation - dominated with circuit switched voice  Third generation and beyond - dominated by data ITU vision for cellular services Traffic planning in heterogeneous cellular networks of the future takes central stage

8. Florida Institute of technologies Page 8 Description of queuing systems  Queuing systems oMathematical abstraction oUsed to develop the traffic analysis and planning methodology  Elements of a queuing system osource population oqueue oservers odistributions of interarrival times, service times, queuing discipline, etc. Outline of a queuing system  Queuing system – cell site  Servers – channel resources – trunks  Population – users connecting to cellular network

9. Florida Institute of technologies Page 9 Source population  Consists of all users that are eligible for service  The most important property - size oinfinite population - arrival rate does not depend on the number of users in the system ofinite population - arrival rate depends on the number of users in the system oif 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  Traditionally cellular systems are dimensioned for a good performance during the busiest hour Example of a call stats benchmarking map

10. Florida Institute of technologies Page 10 Arrival rate and interarrival times  Arrival rate - number of service requests per unit time  The ability of the queuing system to provide effective service depends on distribution of arrival rates  Standard way of specifying arrival rate is through probability density function of interarrival times Example: 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.

11. Florida Institute of technologies Page 11 Service time (call holding time-CHT)  Service time-period of time that the resource is allocated to individual user  Usually specified through its distribution  Most commonly, CHT is exponentially distributed Example: Duration of CHT at a cell Exponential distribution T – average call holding time Note: Exponential distribution is a good model for demand generated by humans (voice, SMS, email,..)

12. Florida Institute of technologies Page 12 Average resource occupancy - traffic in erlangs  Erlang - unit for measuring of traffic intensity  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  Total traffic carried by service facility cannot exceed number of servers Example Definition Resource occupancy time Averaging time [erlang]

13. Florida Institute of technologies Page 13 Alternative interpretation of erlang traffic  Traffic in erlangs = average number of simultaneously occupied servers  Can be measured easily oregular poling of service facility and logging the number of occupied resources Traffic in erlangs for multi-server system Sum of times during exactly n out of C servers are held simultaneously Number of servers Averaging time Example of traffic measurements. Averaging time is 60 min. Poling time is 1 min.

14. Florida Institute of technologies Page 14 Offered, carried and lost traffic  Offered traffic - traffic that would be served if the number of resources is unlimited  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 Relation between offered, carried and lost traffic Note : Communication systems are frequently designed to operate with a certain percentage of lost traffic