Scribd
Upload
179 views21 pages

Trafic Engineering

Uploaded by

dipu ghosh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
179 views21 pages

Trafic Engineering

Uploaded by

dipu ghosh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 21
Traffic engineering provides the basis for the analysis and design of tele- communication networks. In Chapters 2-4 and 6 we calculated blocking probabilities due to the nonavailability of switching paths. The calculations were based on a quantity that, specified the fraction of the time for which a subscriber line may be busy. In practice, the situation is much more complex It is not only the switching elements but also many other common shared sub- systems in a telecommunication network that contribute to the blocking of a subscriber call. In a telephone network, these include digit receivers, inter- stage switching links, call processors and trunks between exchanges. The load or the traffic pattern on the network varies during the day with heavy traffic at certain times and low traffic at other times. The task of designing cost effective networks that provide the required quality of service under varied traffic conditions demands a formal scientific basis. Such a basis is provided by traffic engineering or teletraffic theory. Traffic enginecring analysis cnables one to determine the ability ofa telecommunication network to carry’a given traffic at a particular loss probability. It provides a means to determine the quantum of common equipments required to provide a parti- cular level of service for a given traffic pattern and volume. x, pak 5 beatfte aft Network Traffic Load and Parameters a a In a telephone network, the traffic load on a typical working day during the 24 hours is shown in Fig.8.1. The originating call amplitudes are relative and the actual values depend on the area where the statistics is collected. The traffic pattern, however, is the same irrespective of the area considered. Obviously, there is little use of the network during 0 and 6 hours when most of the population is asleep. There is a large peak around mid-forenoon and mid-afternoon signifying busy office activities. The afternoon peak is, however, slightly smaller. The load is low during the lunch-hour period, i.e. 12,00-14.00 hours. The period 17.00-18.00 hours is characterised by\low™ traffic signifying that the people are on the move from offices to their residences. The peak of domestic calls occurs after 18.00 hours when persons teach home and reduced tariff applies. In many countries including India, 19 27% 214 ‘Traffic Engineering i ing whi iff applies has been changed to begin d during which the reduced tariff app! b rasta 18.00 hee and one may expect the domestic call patterns also to change accordingly. During holidays and festival days the traffic pattern is different from that shown in Fig. 8.1. Generally, there is a peak of calls Number of calls in the hour | 1 = o1 4 7 10 3 (16 19 23 Hour of the day —-————> Fig. 8.1 Typical telephone traffic pattern on a working day. around 10.00 hours just before people leave their homes on outings and another peak occurs again in the evening. x/ Ina day, the 60-mint ¢ interval in which thd hestjs called the busy hour (BH). In Fig. 8.1, the 1-hour period between 11.00 and 12.00 hours is the busy hour. The busy hour may vary from exchange to ex- change depending on the location and the community interest of the subscribers. The busy hour may also show seasonal, weekly and in some places even daily variations. In addition to these variations, there are also unpredictable peaks caused by stock market or money market activity, weather, natural disaster, international events, sporting events etc. To take into account such fluctuations while designing switching networks, three types of busy hours are defined by CCITT in its recommendations E.600: aff Busy Hour: Continuous 1-hour period lyi : a i lying whol 1 Vv concerned, for which the traffic volume or the numberof tiie nner isgreatest. catest. Wf, Peak Busy Hour: The busy hour each day: itusually varies from _day Lit to day, or over a number of days, “4 of ‘Time Cgnsistent Busy Hour: The 1. or For ease of records, the busy hour is taken tw commence on the hour or* half-hour only. Pees e Not all call attempts materialise into actual conversations for a variety of reasons such as called line busy, no answer from the called line, and blocking jn the trunk-groups oF the switching centres. A call gttempt is said to be successful or completed if the called party answ SGall-com (CER) is defin of call-attempt3) ace ; nd which ts an important parameter in deciding the tasaniells mon control or a stored program control system of an ref angoPEME CCR parameter is used in dimensioning the network gapacity. Networks are usually designed to provide an overall CCR of over 0.70. A CCR value of 0.75 is considered excellent and attempts to further improve the value is generally not cost effective (A related parameter that is often used in traffic engineering calculations is T e busy hour calling rate which is defined ashe average numberof calls originated by a subscriber i how ~~ 4 ee as {i An exchange serves 2000 subscribers. If the average BHCA is 10,000-and the CCR is 60%, calculate the busy hour calling rate. Solution sof Average busy hour calls = BHGA’x CCR =6000 calls ABusy hour calling rate = —averape busy hour calls G) total number of subscribers aw The busy hour calling rate is useful in sizing the exchange to handle the peak traffic. In.a rural exchange, the busy hour calling rate may be as low as 0.2, whereas in a business city it may be as high as three or more. Another useful information is to know how much of the day's total traffic is carried during the busy hour. This is measured in terms of day-to-busy hour traffic ratio wish bea ef ulin hea cinch day. Typically, this ratio may be over 20 for a city business area and around six or seven for a rural area. ‘ The traffic load on a given network may be on the local switching unit, interoffice trunk lines or other common subsystems. For analytical treatment in this text, all the common subsystems of a telecommunication nctwork 2 collectively termed as servers, In other publications, the term link or tf ® wyflis used. {The traffic on the network may then, be measured in terms of occupancy of the servers in the network, Such a measure is called the & intensity which is defined as, WA ag = Petiod for which a server is gecupied @0 é © = {otal period of observation 276 Nae none * Vo pare eeten ti h Ao is obvious! i f observation is taken as one jour, Ao 1s 'y Genel, the petiog ed uf to honour the Danish telephone engi- Seer use. work in traffic engineering, His paper 'A K, Erlang, who did pionecrin on traffic theory published in 1909 is now regarded as a cl: rversi® uy entire observation) Traffic int ity may also be specified over a number of servers. . ‘XAMPLE 8.2 In a group of 10 servers, cach is occu cd for 3 Fates in an obscrvation interval of two hours. Calculate the traffic carried by the group. Solution — occupied duration total duration - =p 7025 E Traffic carried per server = ¥& Total traffic carried by the group 10 x 0.25 Erlang measure indicates the average number of servers ‘occupied and is useful in deriving the average number of calls put through during the period of observation. fe 8.3 A group of 20 servers carry a traffic of 10 erlangs. If the av¢rage duration of a call is three minutes, calculate the number of calls put through by a single server and the group as a whole in a one-hour period. Solution Ufrrattic per server = 2 = 058 i.e. a server is busy for80 minutegin one hour. lumber of call c au) YN calls put Yarough by one server = z= 10 calls Viotal number of calls put through by the group = 10 x 20 wR = 200 cals sure of traffig intensity | nt day networks which a better measure to use ds (CS) and eal ity. We may note is valid only in support voice, data and p for representing the traffic inte; minutes (CM) are also used that * Network Traffic Load and Parameters 277 GEE 36ces = 3600¢5='60cM) * EXAMPLE “gf A subscribér makes three phone calls of three minutes, ur minutes and two minutes duration in a one-hour period. Calculate the subscriber traffic in erlangs, CCS and CM. Solution _ — PFsubscriber trafic in erlangs = Pusyperiod _ 34442 total period 60 Traffic in CCS = G+44Hx 60 - 540 _ 700 = t00 = 5.4CCS Vrrafficin CM = 344 +2=9CMy, | Whenever we usc the terminology “subscriber traffic” or “trunk traffic”, we mean the traffic intensity contributed by a subscriber or the traffic intensity on a trunk. As mentioned above, traffic intensity is a call-time product. Hence two important parameters that are required to estimate the traffic intensity or the network load are @ Average call arrival rate(©) © Average holding time per call (én) We cau: then express the (gad offered) to the network in terms of these parameters as cal sin - aa, C and tq must be expressed in Jike time units, For example, if C is in number of calls per minute, fh must be in minutes per oh Wer a 20-minute observation interval, 4g) I duration of the calls is 4800 seconds. Cale: offered to the network by the subscribers and the average subscriber traffic. i Solution — W/Mean arrival rate C = 40/20 = 2 calls/minute mmiver Mean holding time ty = as 2 minutes/call Therefore, * Offered load = 2x 2 =@B) & Average subscriber traffic = 4/40 =[0.1E] mo In the above discussions, we have calculated the traffic in iwo we itd based on the(traffic generated by the subscriberd.and the other base a oe | @bservation of busy servers) in the network. It is possible that the t: Se CU What 15 Gaade ey 9 ! Sea _ 1S 0-0022 218 ‘Traffic Engineering generated by the subscribers sometimes exceeds the network capacity. There are (wo ways in which this overload traffic may be handled: The overload tratfic may be iejected without being serviced or held in @ quetie nil th network facilities become available. tn the first case, the calls are lost and in the second case the calls are delayed. Correspondingly, two types of systems, called loss systems and delay systems are encountered. mnventional snare opines bap oe he ee ee traffic gonditions a user call is blocked and is not serviced unless the user makes a retry. On the other hand, @perator-oriented manual exchanges can be cohsideréd as delay systemg A good operator registers the user request and establishes connection as soon as network facilities become available without the user having to make another request. In data networks, circuit-switched networks behave as loss systems whereas store- (S&F) message or packet networks behave as delay systems. delay systems behave as loss systems. For example, in a S&F n queue buffers become full, then further requests have to be rejected. © basic performance Parameters for a loss system are the grade of ae and the blocking probability, and for a delay system, the service of delay exceeding a certain limit, or and-forward In the limit, etwork if the wh? Grade of Service and Blocking Probability In loss systems, the traffic carrie actual traffic offered to the net: ‘eigeted bythensworkisaninderof the quality at he; Service offered bythe» his is termed grade of service (GOS) and j is defi i So offered waite Cae ae ined as the ratio of d number of calls generated by the Users and the ae Ea i i a ‘This is given by Eq. (82). On the other hand he ses sins itue per cal) fi and is the aver servers in the network as given by Eq. (8.1). 4 = T88€ Occupancy of the 811). Accordingly, Gog ic ae, A-Ay ys is given by GOS= (8.3) where VA = olfered traffic @-tK Ao = carried traffic VA — Ao = lost traffic Grade of Service and Blocking Probability © 279 ‘Thersmaller:the value of gradél6f service, the better is the service. The recommended value for GOS‘in India is 0.002 which means that two callsiin 1000 calls or one call in every 500 calls may be | Usually, every common subsystem in a network has an associated GOS value. fhe GOS of the full network is determined by the highest GOS valuc of the subsystems in a simplistic sense) A better estimate takes into account the connectivity of the subsystems, such as parallel units. Since the vol fi Bes time passes by, the GOS value of a penronk deteaiggaies with time. In order to maintain the value within reasonable limits, imitially the network is sized to have a much smaller GOS value than the recommended one so that the GOS value continucs to be within limits as the network traffic grows. The _blocki is defi lity the Servers in a'systenv are bus) When all the servers are busy, no further traffic can be carried by the system and the arriving subscriber traffic is blocked. At the first instance, it may appear that the blocking probability i# the same measure as the GOS. The Probability that all the servers are busy may well represent the;fraction of the calls log>which is what the GOS is all about. However, this is generally not truer example, in a system with equal _ umber of servers and subscribers, the GOS is zero as there is alwaysaserver available to a subscriber. On the other hand, there is a definite probability that all the servers are busy at a given instant and hence the blocking . Probability is nonzero. The fundamental difference is that theGGOS is a measure from the subscriber point of view whereas the blocking probability is a measure from the network or switching system point of view) GOS is arrived at by observing the number ol sejected subscriber calls , whereas the blocking probability is arrived at by observing the busy servers in the switchs. ing system. We shall show later in this chapter, through analysis carried out on loss systems, that GOS and Pp may have different values depending upon the traffic characterisation model used. In order to distinguish between these two terms clearly, GOSsis:calle: ility and the ing probability is called tim io “__Anthe case of delay systems, the traffic carried by the net as the load offedéd to the network by the subscribers, traffic is queued, all calls are put through the network when the network facilities become available, GOS as defined above is not meaningful in the case of delay systems since it has a valuc of zero always. Theiprobability thata i i work is the same fm ility, is a useful measure, as far as the delay systems are concerned, Ifthe off cred load or the input rate f traffic far exceeds the network cai acity, then the queuc fength: m ery large and the calls experience undesirably lon, delays, Under such is possible that there are some lay systems to the unstable region of he system back to stable region of Sputts of traffic that tend to take the de! Operation. An casy way of bringing t 280 ‘Traffic Engineering operation is to make it bchave like a loss system bee ie reas ‘cleared to an acceptable limit. This technique of m: ig ion i control. . as ea tse the term GOS in a broader sense, covering herchines systems and delay systems and representing both loss and ony Pro pabilties in the respective systems, In this text, we USC the term GOS int a = ventional sense to represent the loss probability only as define q. (8.3). In more recent times, a new term called quality of service (QOS) Js being used and it is considered to be more general than the GOS and includes other factors like ete. It is important that the reader be warned of the fact that these terms are used loosely by many writers and the reader will do well to verify what exactly an author means by different terms. It is essential to recognise that there are two different important performance measures, one obtained by observing the subscriber traffic and the other by observing the network behaviour. The terms that are used synonymously in this text from the subscriber and network viewpoints are a Subscriber viewpoint: #GOS = call congestion = loss probability Network viewpoint: ‘¥Blocking probability = time congestion Modelling Switching Systems 7 4 A telecommunication network carries traffic generated by a large number of individual subscribers connected to the network. Subscribers gciverate calls ion by the subscribers and therefore the ching systems in it can be described as a dom process or a stochastic Brocess is one in which S are call ‘d random va ables. Th i © indexed function of se eandien variables. It is generally possible to characterise the Conor MOFE Fandom ‘ qj to acterise the behavi ara process BYsome Satistical properties and thus oe elm performan| a certain probabilityoThe telenn i © futur stochastic process where the numby atid Qualities'as a and the number of simultaneously i recisely estimate reas Te eat a given ena the number gt ee Yeti i subscribers Ra ee Wstant of time but a predicts taneously active certain proba L ie i sure 82 shows typical fluctune can be made with ‘ simultancous ca! hour: period, e tons in the number 0 random process. Pattern signifies a typical i Modelling Switching Systems 281 260 +- 240 + Number of calls ——> a v e & s 8 + + t a 8 i + at 0 s 10 15 20 25 30 Period (min) —> 2. Typical fluctuations in the number of telephone calls. The values taken on by the random variables of a random process may be discrete or continuous. In the case of telephone traffic, the random variable representing the number of simultaneous calls can take on only discrete values whereas a random variable representing temperature variations in an experiment can take on continuous values. Similarly, thé time index of the random variables can be discre'e or continuous. Accordingly, we have four different types of stochastic processes: Continuous time continuous state + Continuous timediscrete state ¥ Discrete time continuous state A discrete state stochastic process is often called a chain, __ Statistical properties of a random process may be obtained in bwo ways: cither by observing its behaviour over a very long period of time, or by observing simultancously, a very large number of statistically identical random sources at any given instant of time.*The stat al propertics obtained using the first method are known as time statistical parameters, na hr Birth-death Processes A de we apply the restriction that the state transitions of a Markov chain can r onl the adjacent states, then we obtain what are known as birth- eath (B-D) processes) It is customary to talk of popul ; ee 4 ion in a B-D process. The number in the population is a random variable! Ae roprésents.the state value of the process. The B-D process moves from its stat " . ate k to state k — Lif a death occurs or moves to state k + 1, if a birth Occurs; it stays in the same state if there is no birth or death during the time period under consideration. This behaviour of the birth-death process is depicted in Fig, 8.4. Birth-death processes are very useful in our analysis of telecommuni- cation networks. Modelling Switching Systems 285 Attimer + Af Death No change process where thesnu (busy servers\repres e pop irth and a call.termination implies'a death. In order to analyse a B-D process, we shall choose a time interval) small enough such that -~> aff there can almost be only @ne stateitransitiod in that interval, & there is only Bpeanuval orfone- Termination but not both, and 4. there may be fgo-arrivallor fermination leaving the stateuunchanged in the time interval At: We further assume that _~ @ the probability-of an Grivallor fermination)in a particular interval is independent of what h ned inthe earlier time’ intervals and | erval At. ' a) With these assumptions we now proceed to determine the dynamics of a switching system modelled as’a birth-death process. Let ¥ P(t) = the probability that the system is in state k at time ¢, Lc. k Vo Ay = galllarrival ra ¢ in'state® VJ My, = Galltermination rate in stat&®) 286 Traffic EngineennnB ime intervaKQD: Then, we have the following probabilitiesrimthe AVP {exactly one arrival] et tly one termination| VP [exactly o Hat ay ic YP [noartival] YP {no termination] =1-“AL ; Finding-thessystemn'instate@st time [GRAFT is given by the Probability of findi equation vy Py(t + Ai) = P= 1A - 14 + Py 4 OMe + AE + (1A ANG — Ak ANP, () (The first term on the right hand side represents the possibility of finding t timeand a birth ora call request occurring during the interval@ ¢ + Ad) The possibility of finding the system in state-k + Lat time ¢ and a death or a call termination occurrin during the interval t,t + At) is given by the'second term, The last term represents the no arrival (8.5) and ignoring the second and no termination cas¢) Expanding equation order(&rterm, we get ! ae Py(t + AN) = Py (Dlg = At + Pa 21 OM + 144 = Ag + HPEOAL + PO Z wey? (8.6) 65) she system in stat Py Ay + Pe OK +1 Py(t + At) - PO yaaa ae — Ax + H4)PE( + (8.7) In the limit{&7> Owe get , j . a) _ p js ar k= 1 + Pic OM +1 ~ Akt HP) (88) This.is the @ifferentia i i sof ‘ ial equatiog) governing the dyna - Equa 8.8) applies for all values eee a Re alle sc progress, there. can-be-no termination of a cal : II. In oth eT i other words, ur oe ere can be mo,state with as the state value. Therefore, for k= 0, Eq. (88) modifies as : , aPo(t) Y a = Pater ~ AoPol : “ w Be v ‘While the above equations actually give [the rate of change of state probabilities} we are often concerned with the steady state operation of the networks. Under steady state coriditions, the state probabilities reach an Incoming Traffic and Service Time Characterisation * 287 maxalue and do “ ie. Py) = P(t) = (P,)For example, when commercial offices start unetioning’s ound hows, and qui ilities stabili ec. Ui) conditions, we have asst these Py) m of =0 ¥ rT it . » and the B-D process becomes f{atignary) Therefore, the steady state equa- tins of a B-D process are Pe-tAk-1t Pea ite e1— Ak +My) = 0 forkET] (8.10) \U Py fy — AgPo = 0 for 5 ae Oe 8.11 = gti cn (8.11) It may be noted that havik switching system is y Eqs. (8.10) and (8.11), when-the-system-is. modelled as a B-D proce: ; : 8.4 Incoming Traffic and Service Time Characterisation We have scen in Section 8.3 that a B-D process is useful in modelling telecommunication networks. Traffic in a telecommunication network is the aggregate of the traffic generated by a large number of individual subscribers connected to the network.(Subscribers gencrate calls in a random manner and hence the telecommunication traffic is characterised as a random pro- “cess, Whenever a subscriber originates a call, he adds one to the number of calls arriving al the network and has no way by which he can reduce the number of calls that have already arrived: We are thus in need of a model that describes an originating process. Interestingly, this Process can be treated as a special case of the B-D process in which the death rate is equal to zero. In other words, there is no death occurring in the process. Such a process is known as a renewal proce: i ure bith pepe i the sense that it can only TE eon a the ime goes by and cannot the population by itselt(The equations gover! B can be casily afrived a We thus have aPy(t) dt 8.13) AP) s ¢ ao ~AgPo(t) fork =0 Steady state equations are not relevant in a renewal process where we are counting the number of call originations or births. Let us say that we start Pg Ay 1 AP aM) For kz 1 (8.12) u 288 ‘Traffic Engineering the observation at the time = 0. As soon asa birth occurs, say at timet = hy it is impossible to find the system in state 0. Similarly, after i births, it is not possible for the system to be in state i-1, i-2ete. Therefore, our concern is only in the dynamics of the system to answer questions like: What is the probability that there are k birth: In Eqs. (8.10)-(8.13), thesbintherateris=dep If yum a.constanubirthcrated whichiisindependent of the:state system, a ic issonsprocess? The governing equations of a s in a given time interval ¢? Poisson process are . FO = AP, 21) -APA() fork 1 7 (8.14a) a = -AP) fork=0 (8.14b) In order to solve these equations, we have to assume’ certain | vonditions, e.g. at time[_= Othe system is in state zero, i.e. no births have taken place, We then have 4 _ fi fork =0 Px(O) = {i fork #0 With these conditions we get the solution for Eq. (8.14b) as Pol) = (8.15) From Eq. (8.14a) and (8.15) we obtain, fork = 1) vip = 200 EE Solving this equation we get AM P(t) = atew* For = 2the solution is ~ 2 - at py = OE En og By induction we write the general solution as =a Anke 8.16 gate ry = OF ee T Equation (8.16) is,the most celebrated Poisson artiv at The equation expresses the probability of finding the system with k members in the population at time ¢. In other words, it represents the probability of k arrivals in-the time interval t. Equation (8.15) represents the probability of Incoming Traffic and Service Time Characterisation 249 zero arrival in a given time interval which is nothing but the probability distri- bution of interarrival times, i.c. the time that clapses between two arrivals. Thus, in a Poisson process, the intcrarrival time is exponentially distributed, This should be so, as the Poisson process is a Markov process which demands, as described carlicr, that the interstate transition times be expo- nentially distributed. EMME x fo A tural telephone exchange normally experiences four all originations’ per minute. What is the probability that exactly eight calls Occur in an arbitrarily chosen interval of 30 seconds? Solution \ 3 “fb #4 = WiScalls per second 2 A= When [= 30)s, [ir = 2\ Therefore, the probability of exactly eight arrivals is given by 8-2 VFO = = In Example 8.6, cight arrivals in 30 seconds represent a four-fold increase in the normal traffic. The solution shows that the probability of such event occurring is very low. The reader would appreciate that such calculations are useful in sizing a switching system. It is important to recognise that the assumption of an arrival rate inde- pendent of the state of the system in a Poisson process implies that we are dealing with infinite number of sources or essentially constant number of sources. If a number of arrivals occur immediately before any subinterval in question, some of the sources become busy and cannot generate further requests. The effect of busy sources is to reduce the average arrival rate unless the source population is infinite or large enough so as not to be affected significantly by the busy sources, hxamre df A switching system serves 10,000 subscribers with a traffic intensity of 0.1 E per subscriber. If there is a sudden spurt in the traffic, increasing the average traffic by 50%, what is the effect on the arrival rate? : Solution Number of active subscri ers durin, ff normal traffic = 1000, (6) increased iratfic 1500. Number of available subscribe! for generating new traffic during GP normal traffic = 9000.6) increased traffic = 8500 mormal tretticl se ’ , 500 % Change in the arrival rate = p00 * 100 = G5") 20 5 Blocking Models and Loss Estimates In Section 8.1, we pointed out that depending on the way-in which overflow traffic is handled, telecommunication systems may be classified: as loss systems or delay systems. The behaviour of loss systems is studied by using blocking models and that of the delay systems by using queuing models. In this section, we analyse loss systems; and in Section 8.6 we deal with delay systems, . From the discussion in Sections 8.3 and 8.4, it is apparent that we are concerned with three aspects while dealing with the analysis of the telecommunication systems: 1. Modelling the system 2. Traffic arrival model 3. Service time distribution We model the system as a B-D process, the arrival as a Poisson process, and the holding time as an exponential or constant time distribution. In loss Systems, the overflow traffic is rejected. In other words, the Overflow traffic experiences blocking from the network, What happens to the overflow traffic that is rejected? There are three ways in which overflow traffic may be handled: Yer Tost Calls Cleared System with Infinite Sources wea Sle Gbsi bed We-firstanalyse an LCC system assuming infinite number of subscribers. The assumption permits us to use Poisson arrival model for the traffic. The arrival rate is independent of the number of subscribers already busy and remains constant irrespective of the state of the System.Lost calls cleared madel is vs ited to the study of the behaviour of tu ik Usually, there are many trunk groups emanating from a switel terminating on adjacent switching offices. Whenever a direct between two switching offices is busy, it is possible to divert th other switching offices using different trunk groups. In this way, the blocked calls in one trunk group are cleared via other trunk groups. In the context of subscriber calls, the LCC model assumes that a subscriber on hearing the engaged tone, hangs up and waits for some length of time before reattempling. He does not realtempt immediately or within a short time, Such calls are considered to have been cleared from the system and the reattempts are treated as new calls. The LCC model is used as a standard for the design and analysis of telccommunication networks in Europe, India and other countrics that adopt Europcan practices, es The LCC model was first studied comprchensively and accurately by AX. Erlang in 1917. The main Purpose of the analysis is to cstimate the blocking probability, and the grade of service. Using Eq. (8.2), wevean oxy FOCESS as (8.28) A=). ing office and trunk group he traflic via 294 Traffic Engineering rage Poisson call arrival rate. Itbe may be recalled that the : s the average call arrival rate irrespective of the arrival distribution, whereas A represents the Poisson arrival rate here. When all the servers in the system are busy, any traffic generated by the Poisson process is rejected by the system. Since the overflow traffic is lost, as far as the network is concerned, there is a different arrival rate which we call effective arrival rate. We denotc the mean effective arrival rate as Co, and the effective arrival rate in state ias Cj. The system : said to be in state j when sin tl , j servers arc busy.A ‘entireincoming trafficisicarried by the netw pork and ‘whenall'theservers are the network is fic i Such a traffic on known as Erlang traffic or pure chance traffic of type 1, In this case, WC have wherc A is the ave! quantity C in Eq. (8.2) represent VG =A fordsicR, VCR=0 where(Ais the number of servers in the system. The mean effective traffic rate Cois calculated as ° R-1 C= > Fi i=0 where(P)is the probability that the system is in state i. The system can be in any one of 0, 1,2 ... R states. Therefore, we have , Po t Py t Pr tin + PR=l (8.29) We may now write the expression of Co as . Co =A(Po + Py tie + Pri) = AG =Pr (8.30) The mean'traffic'carried by the network is given by Ao = Coln (8.31) = ACL Pr)th “ : (8.31a) By Eq.(8.28), this reduces to Ag = A(- Pr) or A-Ay Pp= (8.32) The blocking probability ®rjis the same as the probability that all the servers are busy, ie. Pi® Therefore, from Eq ‘ 15. (8.3) and (8,32), we conclude that cosmtn f/f Bre Mf for LCC model where the traffic arrival i ve ral al is characterised by Poiss cess We now procecd (0 , eulate the blocking probability, Fat this purpose we perform the steady stale analysis of the 1-D process ‘characterising te Blocking Models and Loss Estimates LCC model. We have already discu: the call termination process. A cons here. The call terminati busy servers in the system, are busy, more calls are likely to terminate ii a given termination rate will be higher. Hence, the termination -rate-may be con. (=H) for gsksR : OS oe * where JH = mean call termination rate = ty Vy = call termination rate in stat 295 ssed the arrival process and we now see tant death rate may not be appropriate rate is likely to be dependent on the number of state of the system. Ifa large number of servers ic and the call (8.33) Substituting the values of birth and death rates in Eqs. (8.10) and (8.11), we have Poa rdt Peg Mkt 1) — + MPL =0 Using Eq, (8.28), we get AP, + KP, APs -1 v Pea res fork >0 SP, = APo fork =0 For[k = I] we have AP, + Py APo 2 are Substituting for Pj from Eq. (8.35), 2 AP era Similarly for{k = 2) we have AP, + 2P)- AP, _ AP, A*Po From Eqs. (8.29) and (8.36), R A Py t Aly to. Oe = (8.34) (8.35) 296 ‘Traffic Engineering Therefore, VP%o=——> 7 (8.37) LHAt tee ty Forfk_= RJ substituting for(Po)from Eq. (8.37) into Eq. (8.36), we obtain R we Ppa“ : (8.38) A’ LH Ata tet Ry ‘ Erlang B.forn rl 3 The quantity Pr is the probability that all servers are busy in the system and hence is the blocking probability Pp of the system. We have already shown. that for the Erlang traffic which occurs in the LCC model, the GOS and the blocking probability values are equal. Therefore, the value of GOS is given by Erlang B formula. CCITT has adopted Erlang B formula as the standard for estimating the grade of service of a switching system. 27 When R is large and A small, A/R is very small and the denominator in Eq. (8.37) reduces to e*, Equation (8.36) then becomes » (8.39) which is the same as Poisson equation. In this casc, the traffic is Poisson In_ the limit R tending to infinity, we have ~ | RU-A ARe (8.39a) Ri 7 9 oF PaO Thus, the blocking probability tends to zero with Poisson traffic, oh vA Lost ra Cleared System with Finite Subscribers tlang loss formula was derived in Section 8.5.1 under a fundamenta assumption that call arrivals are independent of the number of active callers Such an assumption is justified only when the number of sources is much larger than the number of servers. This may not always be the case in practice. For example,ina three-stage space di subscribers connected to each input matrix is c servers (alternative paths) available in the netw rate to the system is dependent on the-numbe: occupied as the b' ision network, the number ol omparable to the number of ork. In such cases, the arrival t of subscribers who are not usy subscribers do not generate new calls. The traffic in this case is known as Engest traffic or pure chance traffic of type 2. We now derive expressions governing blocking probabilities in such cases. The ¥ ' | , Blocking Models and Loss Estimates 297 blocking, probabilities in finite source systems are always less than those for infinite source systems since the arrival rate decreases as the number of busy sources increases. Let Ag = arrival rate per subscriber number of busy subscribers = total number of subscribers R = number of servers The offered traffic (arrival rate) when the system is in state k is given by Cy = (N-k)A, forks0

You might also like