Traffic Forecasting & Network Planning Lec 04 Kamran Kamran Nadeem Nadeem
[email protected]
Traffic Intensity (1) • a = λT, where –
λ
= number of carried connections per time unit (arrival rate, call rate)
– T = mean mean duration duration of of a connection connection or holdin holding g time time
• Traffi Traffic c inten intensit sity y is a bare bare numbe number, r, but but in in order order to emph emphasi asize ze the context, context, one often often writes writes as its ”unit” ”unit” erlang erlang (E, erl) • Traffi Traffic c inten intensit sity y descr describe ibes s the the mean mean numb number er of of simul simultan taneou eous s call in progress • Instea Instead d of a "conn "connect ection ion"" we may consid consider er rese reserva rvatio tion n of of any resource (trunk, modem etc)
Traffic Intensity (2) • Example – In a local local switch switch the number number of calls calls in an an hour hour is 1800 1800 – The mean mean holdi holding ng time time of a call call is 3 min min – What What is the the int inten ensi sity ty? ?
• Typica Typicall traffi traffic c intens intensiti ities es per per a singl single e source source are are (frac (fractio tion n of time they are being used) – private subscriber
0.01 - 0.04 erlang
– business subscriber
0.03 - 0.06 erlang
– PB PBX
0.1 - 0.6 erlang
• A loa load d of of 90 90 erl erlan ang g is crea create ted d by by a pop popul ulat atio ion n of of some some 2250 2250 - 9000 9000 private private subscri subscriber bers. s.
Traffic Variations (1) • Predi redict ctiive Vari ariati ations – Trends • Growth Growth due to existi existing ng servic services es – Imp Improv rovemen ementt – Low Low tari tariff ff etc etc
• Grow Growth th due due to to new new ser servi vice ces s
– Profiling • Yearly • Monthly • Weekly • Dail Daily y (inc (incl. l. busy busy hour hour))
– Variat Variation ions s due due to extern external al even events ts • Reg Regular ular e.g. .g. Eid Eid • Irre Irregu gula larr e.g. e.g. Ele Elect ctio ion n days days
Traffic Variations (2) • NonNon-pr pred edic icti tive ve Vari Variat atio ions ns – Short te term • Call ar arrivals • Holding titimes
– Long te term • Vari Variat atio ions ns in prof profil iles es • Probab Probabili ilisti stic c natu nature re of traffi traffic c
– Variat Variation ions s due due to extern external al even events ts • Natu Natura rall disa disast ster ers s
• Ordina Ordinary ry theo theoret retic ic traf traffic fic models models are based based on shor shortt term term random variables (most predictive)
Traffic Variations (3)
Busy Hour (1) • For dimens dimension ioning ing we need need an an esti estima mated ted traffi traffic c load load • Teleph Telephone one networ networks ks use busy busy hour hour for dimens dimensioni ioning ng • Busy hour – Continuo Continuous us 1-hour 1-hour duration duration when when traffic traffic is maxim maximum um – What What is the the busy busy hour hour for for a singl single e day X? – What What is the the busy busy hour hour for for a whole whole mont month? h?
• ITU ITU has has two two majo majorr def defin init itio ions ns – Averag Average e Dail Daily y Peak Peak Hour (ADPH) (ADPH) – Time Time Consi Consiste stent nt Busy Busy Hour Hour (TCBH (TCBH))
Busy Hour (2)
Busy Hour (3)
Telephone Traffic Model • Tele Teleph phon one e traf traffi fic c cons consis ists ts of call calls s – a call occup occupies ies one one channel channel from each of of the links links along along its its route route – call characteri characterizatio zation: n: holding holding time time (in time time units) units)
• Mode Modeli ling ng of offe offere red d traf traffi fic: c: – call arrival arrival process process (at (at which which moment moments s new calls calls arrive) arrive) – holding holding time distribut distribution ion (how (how long long they take) take)
• Link Link mode model: l: a pure pure loss loss syst system em – – – –
a serve serverr corre correspo sponds nds to a channe channell the servic service e rate rate µ depend depends s on the averag average e holdi holding ng time time the numbe numberr of serve servers, rs, n, n, depends depends on the the link capacity capacity when all all channels channels are are occupied occupied,, call admis admission sion control control reject rejects s new calls so that they will be blocked and lost
• Modelling of ca carried tr traffic: – traffic traffic process process tells tells the the number number of ongoi ongoing ng calls calls = the numbe numberr of occupied channels
Tele Telep phone one Traf Traffi fic c Proc rocess ess
Packet-level Model - Data Traffic • Data Data traf traffi fic c con consi sist sts s of of pac packe kets ts – packets packets compete compete with with each each other other for the proces processing sing and and transmis transmission sion resources (statistical multiplexing) – packet packet chara character cterizati ization: on: length length (in (in data data units) units)
• Mode Modeli ling ng of offe offere red d traf traffi fic: c: – packet packet arrival arrival proce process ss (at which which momen moments ts new packets packets arrive arrive)) – packet packet lengt length h distrib distribution ution (how long they they are)
• Link Link mod model el:: a sin singl gle e serv server er que queui uing ng syste system m – the serv service ice rate rate µ depend depends s on the the link link capaci capacity ty and and the aver average age packet length – when the the link is is busy, new packe packets ts are buffere buffered, d, if possibl possible, e, otherwis otherwise e they are lost
• Mode Modeli ling ng of carr carrie ied d traf traffi fic: c: – traffic traffic process process tells tells the numbe numberr of packets packets in the the system system (includin (including g both the packet in transmission and the packets waiting in the buffer)
Packet-level Process (1)
Packet-level Process (2)
Data Traffic at Flow Level • In a longer longer time time scale scale,, data data traf traffic fic may may be though thoughtt to consist of flows – A single single flow flow is describe described d as a continu continuous ous bit bit stream stream with with a possibly varying rate (and not as discrete packets)
• Flow low class lassif ific icat atio ion: n: – Elas Elasti tic c flow flows s • transmiss transmission ion rate adapts adapts to to traffic traffic condition conditions s in the network network by by a congestion control mechanism • e.g. transfe transfers rs of digital digital docume documents nts (HTTP,F (HTTP,FTP,.. TP,...) .) using using TCP
– Stre Stream amin ing g flow flows s • transmiss transmission ion rate rate independe independent nt of traffic traffic conditi conditions ons in the networ network k • e.g. real real time time voice, voice, audio audio and video transmissi transmissions ons using using UDP UDP
Flow Level Model - Elastic • Elas Elasti tic c traf traffi fic c cons consis ists ts of of adap adapti tive ve TCP TCP flo flows ws – flow character characterizati ization: on: size (in data data units) units) – the transfer transfer rate rate and the the duration duration of of an elastic elastic flow flow are not fixed fixed but but depend on the network state dynamically
• Mode Modeli ling ng of offe offere red d traf traffi fic: c: – flow arrival arrival process process (at which which moments moments new flows flows arrive) arrive) – flow flow size size distri distribut bution ion (how (how large large they they are) are)
• Link Link model del: a sha shari rin ng sys syste tem m – due to to lack lack of admis admission sion control, control, no flows flows are rejected rejected – the serv service ice rate rate µ depend depends s on the the link link capaci capacity ty and and the aver average age flow flow size – in the model, model, the the adaptatio adaptation n of the transm transmissi ission on rate is is immediate immediate,, and the link capacity is shared evenly (fairly) among all competing flows
• Mode Modeli ling ng of carr carrie ied d traf traffi fic: c: – traffic traffic process process tells the numbe numberr of flows flows in the system system
Elastic Traffic Process
Flow Level Model - CBR • Stream Streaming ing CBR traffi traffic c consi consists sts of UDP UDP flow flows s with with cons constan tantt bit rate – flow character characterizati ization: on: bit rate and duration duration
• Mode Modeli ling ng of offe offere red d traf traffi fic: c: – flow arrival arrival process process (at which which moments moments new flows flows arrive) arrive) – flow duration duration distribut distribution ion (how long they they last) last)
• Link Link mode model: l: an infi infini nite te syst system em – due to to lack lack of admis admission sion control, control, no flows flows are rejected rejected – the servic service e rate rate µ depend depends s on the averag average e flow flow dura duratio tion n – transmis transmission sion rate rate and and flow duration duration are are insensitiv insensitive e to the network network state – no buffering buffering in in the flow flow level level model: model: when the total total transmi transmissio ssion n rate of the flows exceeds exceeds the link capacity, capacity, bits are are lost (uniformly from all flows)
• Mode Modeli ling ng of carr carrie ied d traf traffi fic: c: – traffic traffic proces process s tells tells the the numbe numberr of flows flows in the system system,, and, and, as well, well, the total bit rate
CBR Traffic Process
Assig ssignm nmen entt 01 • Your Your task task is to resear research ch how PTCL PTCL mode models ls its voice and data traffic – Write Write the paper in your your own words – Do not copy/pas copy/paste te from the internet internet or any any other other resource – Give Give extensiv extensive e referen references ces
• Dead Deadliline ne:: Sund Sunday ay,, Nove Novemb mber er 14, 14, 201 2010 0
The End
