Research Article Modelling of Cellular Networks with ...
Transcript of Research Article Modelling of Cellular Networks with ...
Research ArticleModelling of Cellular Networks with Traffic Overflow
Mariusz GBdbowski, SBawomir Hanczewski, and Maciej Stasiak
The Chair of Communication and Computer Networks, Poznan University of Technology, Polanka 3, 60-965 Poznan, Poland
Correspondence should be addressed to SĆawomir Hanczewski; [email protected]
Received 29 September 2014; Accepted 29 January 2015
Academic Editor: Jian Li
Copyright © 2015 Mariusz GĆÄ bowski et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.
Thepaper proposes a newmethod formodellingmultiservice cellular networkswith traffic overflow.Theproposedmethod employsa model of Erlangâs Ideal Grading (EIG) with multiservice traffic and differentiated availability. The fundamental advantage of theproposed method, as compared to other relevant methods, is a major simplification in modelling systems with traffic overflow thatresults from the elimination of the necessity of a determination of the parameters of overflow traffic, that is, the average value andthe variance. According to the proposed method, calculations in the overflow system can be reduced to calculations in a systemcomposed of one grading only. The paper presents the method for determining availability in such a grading that models a systemwith traffic overflow.The results of analytical calculations were compared with the results of simulation experiments. The results ofthe research study confirm high accuracy of the proposed method.
1. Introduction
New technologies used in building mobile networks evolvevery quickly. Not much long ago, UMTS networks (Uni-versal Mobile Telecommunications System) [1, 2] and theirextension HSPA (high speed packet access) [2, 3] were anovelty. Currently, more advanced 4G networks are beingdeveloped. 4G networks operate according to the LTE (longterm evolution) standard [4] that enables data transmissionto be enhanced in a way much exceeding hitherto existingpossibilities of cellular networks. Mobility of users and theexpanding available bitrates for data transmission makeit possible for mobile networks to become universal andefficient multiservice access networks. The above is partic-ularly apparent in highly urbanized areas where GSM net-works (global system for mobile communications) 900/1800,UMTS, and LTE cooperate with one another within cellularnetworks of one operator.
In order to ensure optimum usage of resources of mobilenetworks that belong to a given operator, it is necessary totake advantage of various traffic management mechanisms[5]. One of such mechanisms is traffic overflow performedwith the application of connection transfer procedures (Han-dOver (HO)) [6â8]. Connection handover ensures mobilityof users on the one hand and makes it possible to utilize
available network resources on the other hand. In mobilenetworks, each transfer of a connection (in the network of agiven operator)âeffected by lack of available resources in thecell in which a new call appearsâcan be treated as overflow[9]. In the existing cellular networks, the handover procedurecan be executed between cells of the same system or betweencells that belong to different systems (e.g., UMTS and GSM).
In 2G networks, the traffic overflow mechanism hasbeen introduced to achieve optimum usage of GSM900 andGSM1800 network resources [6]. In the case when GSM900and GSM1800 networks coexist in one area, handover ofconnections is used to control the admission process for newcalls in such a way as to make resources of the GSM1800network to be used first. Such a procedure results fromthe characteristics of both systems. A network operatingin 900MHz frequency range is characterized by large celldiameters (even up to 35 km), as well as high power ofthe transmitted signal. GSM1800 networks, in turn, havesmaller cell diameters and lower power of the transmittedsignal (as compared to GSM900). An application of a biggernumber of GSM1800 cells would increase the involved costsin constructing the network and, at the same time, increaseits capacity. From the point of view of optimal operation ofthe whole of the network, new calls should then be directedfirst to the GSM1800 network and thenâin the case of
Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2015, Article ID 286490, 15 pageshttp://dx.doi.org/10.1155/2015/286490
2 Mathematical Problems in Engineering
the lack of free resourcesâto the GSM900 network. In orderto impose such a scenario for resource utilizationâdespitehigher signal power in GSM900âeach new call that appearsin the GSM900 network is transferred to the GSM1800network. It is only in the case when there are no free resourcesin the network that new calls remain in the GSM900 network(the HO mechanism will have no effect). Such an operationcorresponds to traffic overflow in which traffic from manycells (or just one cell) in GSM1800 is transferred to one cellin GSM900.
The overflowmechanism is also usedwith the coexistenceof 2G and 3G networks in a given area. In this case, callsare transferred from the 3G network to the 2G networkbased on the IRAT HO (Inter Radio Access TechnologyHandOver) connection transfer procedure [10]. Due to par-ticular properties of the WCDMA (wideband code divisionmultiple access) radio interface [10] used in 3G networks,the parameters deciding on the activation of the IRAT HOmechanism are signal power of a given RSCP (received signalcode power) cell and the signal to noise ratio đž
đ/đ0. It is
worthwhile to note that a handover of a connection from the3G system to 2G can have influence upon a change in theparameters of some services; for example, data transmissionspeed in older generations of wireless systems can be lower.In turn, a transfer of a voice connection between 3G and 2Gsystems is not followed by a change in the quality parametersfor this service. The traffic overflow mechanism has beenalso implemented in 4G systems in the form of the IRATmechanism that manages handover connections between 3Gand 2G networks [11].
The phenomenon of traffic overflow in telecommunica-tions is well-known and has been studied since the mid-20th century. The mechanism was used for the first timein hierarchical networks in which resources (historicallycalled link groups (defining network resources as the linkgroup resulted from the structure of the first telecommu-nications networks in which the basic resource was thelink, while the link corresponded to the physical pair ofwires connecting, e.g., telephone exchanges. In the paper, theterm ânetwork resourcesâ will be used to define resourcesof modern telecommunications networks, whereas, in thedescriptions of analytical models of traffic overflow systems,the notion of the group, viewed as system resources expressedin allocation units, e.g., links, channels, basic bandwidthunits [12], etc., is introduced due to historical reasons)) weredivided into two groups: direct and alternative resources.New calls that appear in the network are offered first todirect resources (direct groups) and then, only in a situationwhen there are no free direct resources available, these callsare transferred to alternative resources (alternative groups).Since the nature of the traffic stream that overflows fromdirect resources and then is offered to alternative resources(called overflow traffic) is different than the nature of trafficstreams offered to direct resources, the modelling processfor such networks could not employ the models that wereavailable at the time (e.g., Erlang model [12]). New methodsfor the analytical modelling of networks with single-service(single-rate) overflow traffic were proposed for the first time
in the 1950s. The most important methods, summarized inSection 2.1, include equivalent random theory [13, 14] andFredericks-Hayward method [15]. First methodologies formodelling of networks with multiservice (multirate) traffic,presented in Section 2.2, were proposed in [16â18].
All the methods for modelling traffic overflow systemsthat have been developed so far require first a determinationof characteristics of overflow traffic, that is, the averagevalue of overflow traffic intensity and the related variance.The present paper proposes a simpler analytical method foroverflow systems calculation that is based on the ErlangâsIdeal Grading model [19â22]. The method does not requirecalculations of the parameters of overflow traffic. To deter-mine traffic characteristics of overflow systems it is sufficientto know the value of traffic offered to the system and thecapacity of direct and alternative resources. The researchstudies carried out by the authors proved high accuracy ofthe above approach.
The paper is organized as follows. Section 2 describesthe analytical network models with traffic overflow that areknown from the literature of the subject. Section 3 proposesa new networkmodel with traffic overflow. Section 4 presentsa comparison of the results of the analytical calculationswith the results of the simulations for different scenarios fortraffic overflow in cellular networks. Section 5 sums up thepaper.
2. Review of Research
2.1. Modelling of Single-Service Systems with Traffic Overflow.The first models of networks with traffic overflow wereproposed in the 1950s. The literature of the subject includes anumber of models of full-availability groups (full-availabilitygroup is themodel of a resource with complete sharing policy[23, 24]) with overflow traffic, derived under the assumptionthat call streams and service streams of traffic offered todirect resources [13, 15, 25] have the exponential distribution.The best known methods include equivalent random theorymethod (ERT method) [13, 14] and Fredericks-Haywardmethod [15]. The basis for both methods is a determinationof the parameters that describe the properties of overflowtraffic, that is, the average value đ of overflow traffic (thefirst moment of the probability distribution for the numberof busy allocation units (as the allocation unit we consider,e.g., link/channel in circuit switching [23], bandwidth (bitrate) unit [24, 26, 27], or load factor (in wireless systems withsoft capacity [12])) in the alternative resource with infinitecapacity) and the secondmoment and the related varianceđ2.These parameters can be used for evaluation of âpeakednessâof the overflow stream by the introduction of the peakednesscoefficientđ that is equal to the ratio between the variance đ2and the average value of overflow traffic đ :
đ =đ2
đ . (1)
Mathematical Problems in Engineering 3
A1 A2 A3
V1 V2 V3......
...
...
Valt
R1, đ21 R2, đ
22 R3, đ
23
(a)
Aâ
Vâ
...
...
Valt
R, đ2
(b)
Figure 1: Graphic presentation of the ERT method: (a) overflow scheme for real network and (b) overflow scheme for equivalent system.
In both the ERT method and Fredericks-Haywardmethod, the parametersđ and đ2 are determined on the basisof Riordan formula [13]:
đ = đŽđžđ(đŽ) ,
đ2
= đ [đŽ
đ + 1 â đŽ + đ + 1 â đ ] ,
(2)
where đŽ is the traffic offered to the direct resource and đ isthe capacity of the direct resource.
The most common form of overflow in telecommunica-tions networks is when call streams from a number of directresources overflow to one alternative resource. If we assumethat PCT1 (pure chance traffic of Type 1 (PCT1) [23], i.e.,traffic generated by infinite number of sources, the arrivalrate is constant and independent of the number of occupiedsources (Poisson streams)) streams that are offered to theseresources are statistically independent, then traffic streamsthat overflow from these resources will also be independent.If this is the case, then the parameters of the total overflowtraffic can be determined:
đ =
đ
â
đ =0
đ đ , đ
2
=
đ
â
đ =0
đ2
đ , (3)
where đ is the number of direct resources, đ đ is the average
value of traffic that overflows from the resource đ , and đ2đ is
the variance of traffic that overflows from the resource đ .In the ERT method all direct resources are replaced by
one equivalent resource with the capacity đâ, to which suchequivalent trafficđŽâ is offered that the parameters of overflowtraffic (đ and đ2) from such a resource are the same as thoseof overflow traffic from all direct resources. The parametersđŽâ and đâ of the equivalent resource can be determined on
the basis of the known values đ and đ2 after providing the
solution to the Riordan system of equations [13], written inthe following form:
đ = đŽâ
đžđâ (đŽâ
) ,
đ2
= đ [đŽâ
đâ + 1 â đŽâ + đ + 1 â đ ] .
(4)
Equivalent traffic đŽâ requires đ
â
+ đalt allocation units(Figure 1) for servicing calls with the assigned blockingprobabilityđž.The required capacity of an alternative resourcecan be finally determined on the basis of Erlang-B formula:
đž = đž(đâ+đalt)
(đŽâ
) . (5)
It is worthwhile to mention at this point that in thismethod the so-called overflow scheme was used for thegraphical representation of a network with traffic overflow.The overflow scheme shows in a simplified way and indepen-dently of the network technology the dependencies betweendirect and alternative resources. Such an exemplary overflowscheme, used in the ERT method, is presented in Figure 1.
The other method for modelling of single-service sys-tems with traffic overflow, the so-called Fredericks-Haywardmethod [15], was proposed in the 1980s.The basic foundationfor this method is a decomposition of the alternative resourcewith the capacity đalt, to which overflow traffic with theparameters đ and đ is offered, into đ identical componentresources with the following parameters:
(i) capacity of a single component resource:
đđ=đaltđ, (6)
(ii) traffic offered to a single component resource:
đ đ=đ
đ, (7)
(iii) peakedness coefficient đđfor traffic offered to each of
the component resources:
đđ=đ2
đ
đ đ
=(1/đ)
2
đ2
đ /đ= 1. (8)
The value of the coefficient đđequal to 1 means that traffic
offered to each of the component resource is Erlang traffic,that is, PCT1 traffic. Therefore, the blocking probability foreach of the resources formed as a result of decomposition,described by the parameters đ
đ= đ/đ, đ
đ= đ /đ, and
4 Mathematical Problems in Engineering
đđ= 1, can be modelled on the basis of Erlang-B Formula
[15]. Since this probability is the same in each of the resourcesformed as a result of the decomposition, it is then possibleto adopt that it also describes the blocking probability in aresource after a decomposition with the parameters đalt, đ ,and đ. Ultimately, we get
đž (đ , đalt, đ) â đž(đ
đ,đaltđ, 1) â đž
đalt/đ(đ
đ) , (9)
where đž is the blocking probability determined on the basisof Erlang-B formula.
Equation (9) is a modified Erlang-B formula that takesinto account the peakedness of the overflow call streamoffered to the alternative resource. In traffic theory, theformula is called the Fredericks-Hayward formula. Thisapproach makes use of the Erlang model only, which, inconsequence, is followed by a significant simplification ascompared to the ERT method.
Fredericks-Hayward method and the ERT method thatboth make it possible to determine traffic properties ofsystems that are characterized by the exponential servicetime and service calls generated according to the Poissondistribution triggered further work on single-service systemswith overflow traffic.The author of [28] attempts to approachthe problem ofmodelling resources with overflow traffic withthe assumption of hyperexponential service time, whereassingle-service systems with overflow traffic generated by afinite number of traffic sources (pure chance traffic of type2 (PCT2) (pure chance traffic of type 2 (PCT2); the trafficgenerated by a finite number of sources. Each idle source hasits own arrival rate and the total arrival rate is proportionalto the number of simultaneous idle sources) traffic [23]) areconsidered in, for example, [29â31].
2.2. Modelling of Multiservice Systems with Traffic Overflow.First studies in modelling systems with multiservice trafficbegan in the 1980s [32â35]. Over time, the problem of adetermination of the occupancy distribution and the block-ing probability in the full-availability groupwithmultiservicetraffic with the known value of the parameter đ [36â38]was handled. One of the first methods enabling modellingsystems with multiservice traffic overflow was the methodsthat are based on MMPP processes (Markov-modulatedPoisson process), and proposed in [39, 40], among others.Theaccuracy of these methods is accompanied, however, withhigh computational complexity of the process of the deter-mination of variances for overflow traffic streams. The firsteffective methodology for modelling multiservice systemswith traffic overflow that made it possible to both analyse anddimension such systemswas proposed in [17, 41]. In theworks[17, 41] the traffic overflow scheme presented in Figure 2 wasconsidered. In the considered network with traffic overflow,each of the direct resources đ (đ = 1, . . . , đ) and the alternativeresource were offered a call from the set M = {1, 2, . . . , đ}
of traffic classes. Each direct resource is offered a maximumof đ traffic classes of PCT1 type. Calls of class đ demand đĄ
đ
allocation units to set up a connection.The intensity of PCT1traffic of class đ offered to the resource đ is equal to đŽ
đ,đ .
......
...
...· · ·V1 V2 Vđ
Valt
E1, E2, . . . , Em
A1,1, A2,1, . . . , Am1,1
t1,1, t2,1, . . . , tm1,1
A1,2, A2,2, . . . , Am2,2
t1,2, t2,2, . . . , tm2,2
A1,đ , A2,đ , . . . , Amđ,đ
t1,đ , t2,đ , . . . , tmđ,đ
R1,1, R2,1, . . . , Rm1,1
đ21,1, đ
22,1, . . . , đ
2m1,1
R1,2, R2,2, . . . , Rm2,2
đ21,2, đ
22,2, . . . , đ
2m2,2
R1,đ , R2,đ , . . . , Rmđ,đ
đ21,đ , đ
22,đ , . . . , đ
2mđ,đ
Figure 2: Overflow scheme for multiservice overflow traffic net-works.
In the case when a given resource did not service calls of agiven traffic class, it was assumed that the traffic intensity forthis stream was equal to zero. Traffic streams of individualclasses, not serviced in direct resources, overflow to analternative resource. The blocking probability đž
đ,đ for calls of
class đ in the direct resource đ can be determined on the basisof the Kaufman-Roberts recursion [34, 35]:
đ [đđ]đđ
=
đ
â
đ=1
đŽđ,đ đĄđ[đđâđĄđ
]đđ
,
đžđ,đ =
đđ
â
đ=đđ âđĄđ+1
[đđ]đđ
,
(10)
where [đđ]đđ
is the occupancy probability of đ allocation unitsin a resourcewith the capacityđ
đ allocation units and [đ
đ]đđ
=
0 for đ < 0.The variance of each of call streams is determined
approximatively by carrying out a decomposition of each realresource đ into đ
đ fictitious component resources with the
capacitiesđđ,đ (đđ is the number of traffic classes offered to the
resource đ ).The capacity of the fictitious component resourceđđ,đ
is defined, after [17], as this part of the real resource đđ
which is not occupied by calls of the remaining classes:
đđ,đ = đđ â
đđ
â
đ=1;đ =đ
đđ,đ đĄđ, (11)
where đđ,đ
determines the average number of calls of class đthat are serviced in the resource đ :
đđ,đ = đŽđ,đ (1 â đž
đ,đ ) . (12)
Each fictitious resource will service calls of one class exclu-sively, which will make it possible to apply Riordan formulasin order to determine the following:
(i) the average value of the intensity of traffic of class đthat overflows from the resource đ :
đ đ,đ = đŽđ,đ â đđ,đ , (13)
Mathematical Problems in Engineering 5
(ii) the variance đ2đ,đ
for particular call streams that over-flow to the alternative resource:
đ2
đ,đ = đ đ,đ [
đŽđ,đ
đđ,đ /đĄđ+ 1 â đŽ
đ,đ + đ đ,đ
+ 1 â đ đ,đ ] , (14)
where the quotient đđ,đ /đĄđnormalizes the system to a
single-service case,(iii) the peakedness coefficient for class đ overflow traffic
stream:
đđ=đ2
đ
đ đ
, (15)
where the parameters of the total traffic of class đoffered to the alternative resource are calculated asfollows:
đ đ=
đ
â
đ =1
đ đ,đ , đ
2
đ=
đ
â
đ =1
đ2
đ,đ . (16)
An alternativemethod for a determination of the capacityof fictitious groups is proposed in [18, 42]. In this method,the capacity of the fictitious group đ
đ,đ is determined as such
a capacity that leads to obtaining the blocking value đžđ,đ , on
the assumption of offered trafficđŽđ,đ and given calls of class đ
being equal to 1 BBU:
đž(đđ,đ ,đŽđ,đ )= đžđ,đ . (17)
To determine the blocking probability of calls in theresource servicing multiservice overflow traffic, the Hay-ward method described in Section 2.1 was generalized in[43]. In the case of the multiservice systems, appropriatevalues of peakedness coefficients are introduced [17] into theKaufman-Roberts model [34, 35] that originally describedmultiservice systems without traffic overflow (i.e., systemswith đ
đ= 1 for all đ):
đ [đđ]đalt/đ
=
đ
â
đ=1
đ đ
đđ
đĄđ[đđâđĄđ
]đalt/đ
,
đžđ,alt =
đalt/đ
â
đ=đalt/đâđĄđ+1
[đđ]đalt/đ
,
(18)
where đ is the so-called collective peakedness coefficient.In [17], the authors propose to define the coefficient đ asthe weighted mean of the coefficients đ
đfor individual call
streams [17]:
đ =
đ
â
đ=1
đđđđ, where: đ
đ=
đ đđĄđ
âđ
đ=1đ đđĄđ
. (19)
Formula (19) adopts that the contribution of the peakednesscoefficient for the stream of class đ in the collective peaked-ness coefficientđ is directly proportional to the value of trafficoffered to the alternative resource by the class đ.
Another general approach to the analysis of multiservice,hierarchical networks with overflow traffic were proposedin [16]. It is based on the process in which the overflowtraffic was approximated by ON/OFF andMMPP traffic.Thismethod is characterized by high computational complexitywhich greatly hinders its practical application.
3. New Approach to Modelling ofOverflow Systems
All the methods presented in earlier sections are based intheir operation on the full-availability group model. It isnecessary in these models to determine values of additionalparameters (đ
đ, đ2
đ) describing traffic that overflows to the
alternative resource. In this section we propose a methodfor a determination of the blocking probability in systemswithmultiservice overflow traffic that is based on the ErlangâsIdeal Grading (EIG) model. The principal advantage of theproposed method is that it is not necessary to determine themean value and the variance of overflow traffic. To determinetraffic characteristics of overflow systems it is sufficient toknow the value of traffic offered to the system, the capacityof direct and alternative resources, and the value of the so-called availability parameter.
The proposed method relates to the remark given bythe authors of the book on the life and lifetime achieve-ments of Agner Krarup Erlang [44], who noticed that asimple EIG approximates other nonfull-availability groups(gradings) very well, including those that correspond to thesimplest symmetrical overflow schemes (overflow schemesymmetry is understood here as the identical structureof direct resources, i.e., the same capacity and the sameoffered traffic). The remark of the authors of the book [44]passed unnoticed and has never been addressed by scientificcommunity that dealt with research pertaining to trafficoverflow. The reason for the above is the asymmetry ofoverflow systems, that is, different capacity of direct resourcesand, in consequence, different availability of resources of anoverflow scheme for individual call streams offered to directresources. It was only the construction of the generalizedEIG model [45] with different values of offered traffic anddifferent availabilities for different call classes that madethe problem of approximation of asymmetrical overflowschemes by such groups very interesting. Reference [46]attempts for the first time to apply a model of EIG for theanalysis of simplified multiservice overflow system in whichonly one call class is offered to given direct resources. Thestudy showed that the use of the Erlangâs Ideal Grading tocalculate the blocking probability in networks with trafficoverflow provided the accuracy of calculations comparable toothers available but more complex methods described in theliterature of the subject. Further on in Section 3 a generalizedmodel of multiservice networks with traffic overflow will beproposed in which direct resources can service a numberof traffic classes. To enhance the presentation of the basicassumptions of the proposed method for modelling systemswith multiservice overflow traffic, the basic assumption ofErlangâs Ideal Grading will be outlined in Section 3.1.
6 Mathematical Problems in Engineering
A/3
A/3
A/3
V = 3
1
3
2
d = 2
d = 2
d = 2
Load groups
(a)
A/3
A/3
A/3
V = 3
1
1
3
3
2
2
d = 2
d = 2
d = 2
(b)
Figure 3: Erlangâs Ideal Grading (a) offered traffic distribution and (b) concept of availability [22].
3.1. Erlangâs Ideal Grading. Erlangâs Ideal Grading (EIG) forsingle-service trafficwas defined by A. K. Erlang [19, 47], whointroduced a formula for a determination of the occupancydistribution in the grading in question. EIG, schematicallypresented in Figure 3(a), is described by three parameters:đŽ,đ, and đ, wheređŽ is the average traffic offered to the grading,đ is the grading capacity, and đ is the so-called availability.Availability determines resources that are available for a givengroup of inputs, the so-called load group.The number of loadgroups đ in EIG is equal to the number of possible ways ofchoosing the đ allocation units from their general numberđ:
đ = (
đ
đ) . (20)
Further assumption is that traffic offered by calls is distributeduniformly to all load groups.
Figure 3 shows Erlangâs Ideal Grading with the followingparameters: đ = 3, đ = 2, and đ = 3. Due to the uniformdistribution of traffic with the intensity đŽ to all load groups,traffic offered by each load group is đŽ/3.
In 1993, in [20] a model for Erlangâs Ideal Grading withmultiservice traffic was proposed. Soon afterwards, in [21,45], this model was expanded for the case in which eachclass can have a different value of the availability parameterand then for the case of noninteger values of the availabilityparameter. An exemplary model of themultiservice EIGwiththe capacity đ = 3 with different values of the availabilityparameter (đ
1= 2, đĄ1= 1, đ
1= 3, đ
2= 3, đĄ2= 3, and đ
2= 1)
is shown in Figure 4. The grading under scrutiny was offeredstreams of two traffic classes with the intensities đŽ
1and đŽ
2,
respectively. TrafficđŽ1was distributed uniformly into đ
1= 3
load groups, whereas traffic đŽ2was distributed into đ
2= 1
load group.According to [45], the occupancy distribution in such a
grading can be described by the following formula:
đ [đđ]đ=
đ
â
đ=1
đŽđđĄđđđ(đ â đĄ
đ) [đđâđĄđ
]đ
, (21)
where đŽđis the traffic intensity of class đ traffic, đĄ
đis the
number of allocation units demanded by a call of class đ, đis the number of traffic classes, and đ
đ(đ) is the conditional
probability of transition for a class đ traffic stream in occu-pancy state đ:
đđ(đ) = 1 â
đđ
â
đ„=đđâđĄđ+1
đđ,đđ
(đ, đ„) , (22)
where đđ= đ if (đ
đâ đĄđ+ 1) †đ < đ
đ, and đ
đ= đđif đ â„ đ
đ.
Probability đđ,đđ
(đ, đ„) is the probability of occupancyof đ„ allocation units in a given load group, under thecondition that in the whole of the grading đ allocation unitsare occupied. The probability đ
đ,đđ
(đ, đ„) is described by ahypergeometric distribution:
đđ,đđ
(đ, đ„) =(đđ
đ„) (đâđđ
đâđ„)
(đ
đ)
, (23)
where đđis the availability for calls of class đ expressed in
allocation units.The blocking probability for calls of class đ in the consid-
ered model is determined by the following formula:
đžđ=
đ
â
đ=đđâđĄđ+1
[1 â đđ(đ)] [đ
đ]đ. (24)
3.2. New Network Model with Traffic Overflow. Let usconsider now a new approach to modelling systems withmultiservice traffic overflows. The presented multiservicetraffic overflow scheme shown in Figure 5 will be used todemonstrate the operation of the method. The overflowsystem presented in Figure 5(a) is composed of V directresources and an alternative resource.
The considered network with traffic overflow servicescalls from the set M = {1, 2, . . . , đ} of traffic classes. Eachdirect resource is offered the maximum ofđ traffic classes ofPCT1 type. A call of class đ appearing at the input of resourceđ demands đĄ
đallocation units to set up a connection. The
assumption is that calls of class đ demand the same numberof allocation units in each of the direct resources and in thealternative resource. The capacity of the alternative resourceis equal to đalt allocation units, whereas the capacity of thedirect resource đ is đ
đ allocation units.
The intensity of PCT1 traffic of class đ offered to theresource đ is equal to đŽ
đ,đ . In the case when a given resource
Mathematical Problems in Engineering 7
A1/3
A1/3
A1/3
A2/1
1
2
3
d2 = 3
d1 = 2
d1 = 2
d1 = 2
V = 3
(a)
A1/3
A1/3
A1/3
A2/1
1
12
23
3
d2 = 3
d1 = 2
d1 = 2
d1 = 2
V = 3
(b)
Figure 4: Erlangâs Ideal Grading with different availability (a) offered traffic distribution, (b) concept of availability [22].
does not service calls of a given traffic class, it is adoptedthat the traffic intensity for this stream is equal to zero.Traffic streams of individual classes, not serviced in directresources, overflow to an alternative resource.The alternativeresource can also service traffic that is offered directlyânotonly overflow traffic from direct resources (Figure 5(a)). Theintensity of PCT1 traffic of class đ offered directly to thealternative resource is equal to đŽ
đ,alt.In the proposed method, calculations of the blocking
probability in the overflow system composed of many full-availability groups (both direct and alternative resources)can be reduced to calculations of this probability in asystem that is composed of one group only: Erlangâs IdealGrading (Figure 5(b)). According to the proposed method,the blocking probability in the overflow system is determinedfrom the following dependence:
{[đž1,1, . . . , đž
đ,1, . . . , đž
đ,1] , . . . , [đž
1,đ , . . . , đž
đ,đ , . . . , đž
đ,đ ] ,
. . . , [đž1,alt, . . . , đžđ,alt, . . . , đžđ,alt]}
= đ {[(đŽ1,1, đĄ1, đ1,1) , . . . , (đŽ
đ,1, đĄđ, đđ,1) , . . . ,
(đŽđ,1, đĄđ, đđ,1)] , . . . ,
[(đŽ1,đ , đĄ1, đ1,đ ) , . . . ,
(đŽđ,đ , đĄđ, đđ,đ ) , . . . , (đŽ
đ,đ , đĄđ, đđ,đ )] , . . . ,
[(đŽ1,đ, đĄ1, đ1,đ) , . . . , (đŽ
đ,đ, đĄđ, đđ,đ) , . . . ,
(đŽđ,đ, đĄđ, đđ,đ)] ,
[(đŽ1,alt, đĄ1, đ1,alt) , . . . , (đŽđ,alt, đĄđ, đđ,alt) , . . . ,
(đŽđ,alt, đĄđ, đđ,alt)] , đ} ,
(25)
where đ is the total capacity of Erlangâs Ideal Gradingapproximating the overflow system, đ
đ,đ is the availability
for calls of class đ, offered to direct resource đ , in EIGapproximating the overflow system, đ
đ,alt is the availabilityfor calls of class đ, offered directly to the alternative resource,in EIG approximating the overflow system, đŽ
đ,đ is the traffic
intensity offered to direct resource đ by calls of class đ,đŽđ,alt is the traffic intensity offered directly to the alternative
resource by calls of class đ, đĄđis the number of allocation units
demanded by calls of class đ, andđ is the function that definesthe dependence between the values of blocking of particulartraffic classes offered to the network and the traffic intensityof offered streams and the availability.
At the first stage of defining the function đ that makesa determination of the values for blocking in call streamsoffered to a network with traffic overflow possible, the occu-pancy distribution in the nonfull-availability group (grading)that approximates the network with traffic overflow is deter-mined on the basis of the modified formula (21):
đ [đđ]đ=
đ
â
đ =1
đ
â
đ=1
đŽđ,đ đĄđđđ,đ (đ â đĄ
đ) [đđâđĄđ
]đ
+
đ
â
đ=1
đŽđ,altđĄđđđ,alt (đ â đĄđ) [đđâđĄ
đ
]đ
.
(26)
In (26), đđ,đ (đ) is the conditional probability of transition
for a class đ traffic stream, offered to the direct resource đ , inoccupancy state đ:
đđ,đ (đ) = 1 â
đđ,đ
â
đđ,đ âđĄđ+1
đđ,đđ
(đ, đ„) , (27)
where đđ,đ
= đ if (đđ,đ â đĄđ+ 1) †đ < đ
đ,đ , and đ
đ,đ = đđ,đ
ifđ â„ đđ,đ .
Probability đđ,đđ,đ
(đ, đ„) is the occupancy probability ofđ„ allocation units in resources available for calls of classđ offered to the direct resource đ , determined under thecondition that there are đ busy allocation units in the wholeresources:
đđ,đđ,đ
(đ, đ„) =
(đđ,đ
đ„) (đâđđ,đ
đâđ„)
(đ
đ)
, (28)
where đđ,đ
is the availabilityâexpressed in allocation units âfor calls of class đ offered to direct resource đ .
The value of the parameter đđ,alt in Formula (26) can
be determined on the basis of Formulas (27) and (28), byreplacing the index of direct resource âđ â with the index ofalternative resource âalt.â
8 Mathematical Problems in Engineering
· · ·...
...
...
A1,đ, A2,đ , . . . , Am,đt1, t2, . . . , tm
A1,1, A2,1, . . . , Am,1
t1, t2, . . . , tm
A1,alt , A2,alt , . . . , Am,altt1, t2, . . . , tm
Valt
V1
E1, E2, . . . , Em
Vđ
(a)
· · ·
...
A1,đ, A2,đ , . . . , Am,đ
t1, t2, . . . , tm
A1,1, A2,1, . . . , Am,1t1, t2, . . . , tm
A1,alt , A2,alt , . . . , Am,altt1, t2, . . . , tm
E1, E2, . . . , Em
V
dm,alt
d1,alt
dm,đ
d2,altd2,đd1,đ
d2,1
d1,1dm1,1
(b)
Figure 5: Presentation of the method for modelling systems with traffic overflow with the application of Erlangâs Ideal Grading (a) overflowscheme for real network and (b) equivalent grading.
The proposedmodel assumes that the capacity of ErlangâsIdeal Grading used for the calculations is determined as thesum of the capacities of the resources in the system underconsideration, that is, the capacity of all direct resources andthe alternative resource:
đ =
V
â
đ =1
đđ + đalt. (29)
Such an approach results in the approximating Erlangâs IdealGrading to be corresponding, in terms of its capacity, to theoverflow system (considered group of cells) for which thecalculations are to be made. Moreover, offered traffic to asingle allocation unit is the same in both systems (real andits analytical model).
The values of the availability parameters for individualcall classes are determined in the following way: a call ofclass đ that appears at the input of the direct resource đ canbe serviced only when it can be serviced entirely by thedirect resource đ or, if the resource is busy, by the alternativeresource.Hence, the value of the availability parameterwill bethe sumof availabilities in the direct resource đ (at the input ofwhich the considered call class appears) and in the alternativeresource. In the case when the resource capacity is not aninteger multiple of the number of allocation units demandedby a call of class đ, it is assumed that the availability in thedirect resource will be defined by the number of allocationunits that corresponds to the maximum number of calls ofclass đ that can be serviced in this resource, that is, âđ
đ /đĄđâđĄđ.
Availability in the alternative resource is determined in asimilar way, that is, on the basis of the maximum numberof allocation units that calls of class đ can occupy in thealternative resource:
đđ,alt = â
đaltđĄđ
â đĄđ. (30)
Taking into consideration the availability for class đ calls(offered to the direct resource đ ) in the direct resource đ
and the alternative resource, the total availability will bedetermined on the basis of the following formula:
đđ,đ = â
đđ
đĄđ
â đĄđ+ â
đaltđĄđ
â đĄđ. (31)
Having determined the capacity and the availability thatapproximates the overflow system by the EIG, we are inposition to determine the occupancy distribution in ErlangâsIdeal Grading on the basis of formulas (26) and (27) and then,the blocking probability for calls of class đ:
đžđ,đ =
đ
â
đ=đđ,đ âđĄđ+1
[1 â đđ,đ (đ)] [đ
đ]đ. (32)
The sequence of calculations in the proposedmethod canbe written in the form of the Overflow-EIG method:
Overflow-EIG Method
(1) Define capacityđ of the approximating Erlangâs IdealGrading (formula (29)).
(2) Determine availabilities đđ,đ
for call classes that areoffered to individual direct resources (formula (31)).
(3) Determine availabilities đđ,alt for call classes that are
offered directly to the alternative resource (formula(30)).
(4) Determine the occupancy distribution in ErlangâsIdeal Grading (formula (26)).
(5) Determine the blocking probability in Erlangâs IdealGrading (formula (32)) that approximates the block-ing probability of the overflow system.
Observe that the analytical model of the network withmultiservice traffic overflow (each of direct resources servicesmany traffic classes) proposed in the paper is simplified to
Mathematical Problems in Engineering 9
......
......
...
......
· · · · · ·
A1,1, . . . , Am,1
t1, . . . , tm
t1, . . . , tm t1, . . . , tm
t1, . . . , tm t1, . . . , tm
t1, . . . , tmt1, . . . , tm
A1,i , . . . , Am,i
t1, . . . , tm
A1,i+1, . . . , Am,i+1
t1, . . . , tm
A1,s, . . . , Am,s
t1, . . . , tm
V1 Vi VsVi+1
Valt-T2,2Valt-T2,1
Valt-T3,3
Figure 6: Scheme of three-tier overflow.
the model considered in [46] if each of the direct resourceswill be offered only one traffic stream.
The method for modelling networks with multiserviceoverflow traffic proposed in the paper is presented for thecase of systems in which direct resources are offered Poissoncall streams. It should be stressed, however, that the methodcan be easily adopted for modelling systems with trafficoverflow in which direct resources are offered multiserviceBPP (Bernoulli-Poisson-Pascal) call streams [23]. To achievethat, one can make use of the results of studies devoted tomodelling (determination of occupancy distributions and theblocking probability) of systemswith BPP traffic, such as [48â50].
Theproposedmethod can be easily adopted formodellingof multitier overflow systems. An example of a scheme fortraffic overflow in three-tier systems is shown in Figure 6.Traffic of class đ that overflows from the direct resource đ
đ
(đ = 1, 2, . . . , đ) is directed to one of the two alternativeresources of the second tier, that is, đalt-T2,1 and đalt-T2,2. Ifthe resources do not currently have free allocation units, thenoverflow traffic is directed to an alternative resource in thethird tier đalt-T3,3.
In the considered case, the availability for traffic đŽđ
includes all resources that can service this traffic, and formula(31) can be written in the following way:
đđ,đ = â
đđ
đĄđ
â đĄđ+ â
đalt-T2,đ„
đĄđ
â đĄđ+ â
đalt-T3đĄđ
â đĄđ, (33)
where đ„ is the number of allocation units of the second tierto which traffic from resource đ overflows.
4. Numerical Results for Modelling CellularNetworks with Traffic Overflow
The presented method for modelling multiservice systemswith overflow traffic is an approximate method. To evaluatethe accuracy of the proposed solution, the results of theanalytical calculations of the blocking probability in cellularnetworks with traffic overflow have been compared withsimulation data.The study was conducted formany scenariosof traffic overflow that occur in currently used mobile 2G,2.5G, 3G, and 4G networks. The studies on systems with
traffic overflow, for the scenarios under consideration, werecarried out at the call level. In the simulation model, eachcell was represented as an object with a given capacity andoffered traffic. Depending on the current load of cells towhich calls generated by subscribers were directly offered,these calls were either serviced by the cells or directed to thealternative cell. The analysis of the system at the call levelmakes it possible to consider traffic overflow regardless ofthe technology employed to construct a mobile network. Adetailed description of the methods for modelling mobilesystems at the call level in wireless networks of differenttechnologies is presented in, for example, [12]. The obtainedresults of the blocking probability for individual traffic classesare presented in Figures 8â16 in relation to the value of theaverage traffic offered to an allocation unit of the system:
đ =âđ
đ =1âđ
đ=1đŽđ,đ đĄđ
đalt + âđ
đ =1đđ
. (34)
The diagram for the case of traffic overflow that wasconsidered first is presented in Figure 7. The scheme relatesto traffic overflow between cells of the GSM900 system.Figure 7(a) shows the idea of the handover of connectionsbetween the cells of the considered GSM900 system, whereasFigure 7(b) shows the overflow scheme that was used foranalytical calculations according to the proposed method. Inthe considered system, the area serviced by the base stationwas divided into three sectors with the angle of the coveragearea equal to 120 degrees. To enhance the effectiveness ofthe operation of the network, the base station managed yetanother cell, the so-called omnidirectional cell that servicedthe whole of the area. This cell admits for service only thosecalls that cannot be serviced in one of the three sectors. Thismeans that the omnidirectional cell does not have its owntraffic, that is, calls appearing directly in this cell.
The results for the blocking probability obtained for theconsidered case of traffic overflow (Figure 7) are shown inFigure 8. The assumption was that all cells (sectors) operatedin the same technology (GSM900) and that they serviced onlyvoice connections that demanded đĄ = 1 allocation unit. Thecapacities of the individual sectors and the omnidirectionalcell were identical and were equal to đ
1= đ2= đ3=
đalt = 16 allocation units. Another assumptionwas that trafficoffered to individual sectors had the same value. Since thecapacities of the sectors and the intensities of offered trafficwere identical, the blocking probability for calls appearing ineach cell is the sameâFigure 8 shows then the results for callsoffered to one, selected sector.
Figure 9 presents the results for the blocking probabilityobtained also for the scheme shown in Figure 7, and thistime with the assumption that each cell services two classesof traffic: voice connections (đĄ
1= 1 allocation units) and
data transmission (đĄ2
= 2 allocation units). Similarly asin the earlier case, the capacity of each sector and theomnidirectional cell was đ
1= đ2
= đ3
= đalt =
16 allocation units. Since the capacities of the sectors areidentical; therefore, availabilities for particular call classesin each of the sectors are the same. This means that theblocking probability đž
1,1, đž1,2, and đž
1,3is identical, which
10 Mathematical Problems in Engineering
GSM900
Voice Data
GSM900
V1
V3 V2
Valt
(a)
V3V2V1
Valt ...
......
...
t1 t2 t1 t2 t1 t2
A1,1, t1 A1,2, t1A2,1, t2 A2,2, t2 A1,3, t1 A2,3, t2
(b)
Figure 7: Traffic overflow between cell of the GSM900 system: (a) traffic overflow between cells and (b) overflow scheme (đŽđ„,đŠ
is trafficgenerated by calls of class đ„ in cell đŠ and đĄ
đ„is the number of resources demanded by calls of class đ„).
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Bloc
king
pro
babi
lity
Traffic offered [erl]
Simulationâclass 1Calculation (HAY)âclass 1
Calculation (IVE)âclass 1Calculation (EIG)âclass 1
100
10â1
10â2
10â3
10â4
10â5
10â6
10â7
Figure 8: Blocking probability in mobile system GSM900 withoverflow traffic, only voice traffic.
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Bloc
king
pro
babi
lity
Traffic offered [erl]Simulationâclass 1Calculation (HAY)âclass 1Calculation (IVE)âclass 1Calculation (EIG)âclass 1
Simulationâclass 2Calculation (HAY)âclass 2Calculation (IVE)âclass 2Calculation (EIG)âclass 2
100
10â1
10â2
10â3
10â4
10â5
10â6
10â7
Figure 9: Blocking probability in mobile system GSM900 withoverflow traffic, two traffic classes: voice and data.
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Bloc
king
pro
babi
lity
Traffic offered [erl]
Simulationâclass 1Calculation (EIG)âclass 1Simulationâclass 2Calculation (EIG)âclass 2Simulationâclass 3Calculation (EIG)âclass 3
Simulationâclass 4Calculation (EIG)âclass 4Simulationâclass 5Calculation (EIG)âclass 5Simulationâclass 6Calculation (EIG)âclass 6
100
10â1
10â2
10â3
10â4
10â5
Figure 10: Blocking probability in mobile system GSM900 withoverflow traffic, various capacities of direct resources.
is indicated in Figure 9 as the blocking probability of class1. Similarly, traffic đŽ
2,1, đŽ2,2, and đŽ
2,3is characterized by
identical blocking probability, indicated in Figure 9 as theblocking probability for calls of class 2.
Figure 9 shows the results obtained during the simulationstudy and on the basis of the three analytical methods: theOverflow-EIG method proposed in the paper, generalizedHayward method (HAY) [17] and the method that takes intoconsideration a modification to the way the determinationof the capacity of fictitious groups is executed (IVE) (17). Itis noticeable that for each of the offered traffic classes theproposed Overflow-EIG (EIG) method is characterized bythe greatest accuracy.
Figure 10 shows the results with the assumption that thecapacities of GSM900 cells (direct resources) have different
Mathematical Problems in Engineering 11
GSM1800
Voice Data
GSM900
V1
V4
V3
V2
Valt
(a)
V1 V4V3V2
Valt...
......
......
t1 t2 t1 t2 t1 t2 t1 t2
A1,1, t1 A1,2, t1A2,1, t2 A2,2, t2
A1,3, t1A2,3, t2
A1,4, t1A2,4, t2
(b)
Figure 11: Overflow between cells of systems GSM1800 and GSM900: (a) traffic overflow between cells and (b) overflow scheme (đŽđ„,đŠ
istraffic generated by calls of class đ„ in cell đŠ and đĄ
đ„is the number of demanded resources by calls of class đ„).
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Bloc
king
pro
babi
lity
Traffic offered [erl]
Simulationâclass 1Calculation (HAY)âclass 1Calculation (IVE)âclass 1Calculation (EIG)âclass 1
Simulationâclass 2Calculation (HAY)âclass 2Calculation (IVE)âclass 2Calculation (EIG)âclass 2
100
10â1
10â2
10â3
10â4
10â5
10â6
10â7
Figure 12: Blocking probability in mobile system with overflowtraffic between cells of GSM1800 and GSM900 systems.
capacities that are, respectively, equal to đ1= 16, đ
2= 10,
and đ3= 8 allocation units. The capacity of the alternative
cell is equal tođalt = 16 allocation units.The presented resultsshow that the accuracy of the proposed method does notdepend on the capacities of direct resources. In the case ofa coexistence of the systems: GSM900 and GSM1800 withinthe network of one operator, there is still a possibility forconnections to be transferred between cells. As the examplepresented in Figure 11 shows, the cell GSM900 (with higherpower range than cells of GSM1800) is considered as thealternative resource to which calls from occupied cells ofGSM1800 (operating in the area also covered by the GSM900cell) are directed.
The results of the blocking probability for the considerednetwork with traffic overflow between cells of the GSM1800and GSM900 systems (according to the overflow schemepresented in Figure 11) are shown in Figure 12. The presentedresults were obtained for a network combined of 4 GSM1800cells (direct resources) and one GSM900 cell (alternativeresources). An additional assumption was that the GSM900cell, servicing overflow traffic from 4 GSM1800 cells, had notraffic of its own. In the considered example, each cell servicedtwo classes of call: voice connections (đĄ
1= 4 allocation
units) and data transmission (đĄ2= 1 allocation unit). The
capacity of each of the cells was 32 allocation units. Theresults obtained in the simulation study were compared withthe data provided by the methods. Overflow-EIG generalizedHayward method (HAY) [17] and the method that includes amodification to the way of the determination of the capacityof fictitious groups (IVE) [42]. It is clearly observable thatalso in this case the proposed Overflow-EIG method ischaracterized by the greatest accuracy for each of the offeredtraffic classes. Consequently, in Figures 13â17 the resultsobtained on the basis of Overflow-EIGmethod are comparedwith simulation results only.
The next considered case was a network with trafficoverflow between the systems: UMTS and GSM (Figure 13).These systems differ in terms of available packet transmissionspeed, but assuming that the GSM system is enhanced bythe EDGE, mechanism, redirecting packet transmission tothis system is possible (in the boundary case, only voiceconnections can be transferred). The assumption is thatcalls transferred to the GSM system are first directed tothe GSM1800 system. If there are no free resources, callsare transferred to the GSM900 system. Figure 14 shows theresults for the considered type of overflow, that is, between4 cells of the UMTS system and one GSM1800 cell andone GSM900 cell. The assumption was that the alternativeresource, that is, the GSM cells, did not service its own traffic.
12 Mathematical Problems in Engineering
GSM1800
GSM900
UMTS
Voice Data
V1
V2
V3
V4
Valt1
Valt2
(a)
V2V1 V3 V2......
...
...
......
t1
t1
t2 t3
t2 t3 t1 t2 t3 t1 t2 t3 t1 t2 t3
A1,1, t1 A1,2, t1 A1,3, t1 A1,4, t1A2,3, t2 A2,4, t2
A3,3, t3 A3,4, t3
A2,2, t2
A3,2, t3
A2,1, t2
A3,1, t3
Valt1
Valt2
(b)
Figure 13: Traffic overflow between cells of the UMTS system and GSM1800 and GSM900: (a) traffic overflow between cells and (b) overflowscheme (đŽ
đ„,đŠis traffic generated by calls of class đ„ in cell đŠ and đĄ
đ„is the number of demanded resources by calls of class đ„).
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Bloc
king
pro
babi
lity
Traffic offered [erl]
Simulationâclass 1Calculation (EIG)âclass 1Simulationâclass 2
Calculation (EIG)âclass 2Simulationâclass 3Calculation (EIG)âclass 3
100
10â1
10â2
10â3
10â4
10â5
10â6
10â7
Figure 14: Blocking probability in mobile system with overflowtraffic: overflow between cells of the UMTS system and GSM1800and GSM900.
The capacity of the UMTS cells was equal to 140 allocationunits, whereas the capacity of the GSM cell was 32 allocationunits. Each UMTS cell serviced three classes of calls thatdemanded 1, 4, and 10 allocation units, respectively. Sincethe capacities of all cells were identical, then availabilities forparticular classes of calls offered to each of the cells were alsoidentical. This means that the blocking probability for callsof a given traffic class depends only on the size of demandedresources and not on the cell to which these calls are offered.
Therefore, Figure 14 shows only blocking probabilities for 3traffic classes.
The last considered case of traffic overflow involvestraffic overflow between the systems: LTE and UMTS. Weassume that these two systems are comparable as far asdata transmission possibilities are concerned. Trafficoverflowin this particular case relates to all types of connections(Figure 15). In this scenario, overflow of voice connectionsto the GSM system is omitted. The results of the blockingprobability for calls of particular traffic classes, in the networkwith traffic overflowbetween the LTE and theUMTS systems,are presented in Figure 16. Connections from 3 LTE cellswere directed to one UMTS cell that serviced additionally itsown traffic. The system serviced four classes of calls: voiceconnection class (đĄ
1= 1 allocation units) and three call
classes related to data transmission (đĄ2= 5 allocation units,
đĄ3= 10 allocation units, and đĄ
4= 30 allocation units, resp.).
The capacities of the LTE cells were identical and were equalto đ1= đ2= đ3= 600 allocation units, whereas the capacity
of the UMTS cell was equal to 200 allocation units.The cells 1and 2were offered four traffic classes and the cell 3 was offeredthree traffic classes (Figure 15). Traffic offered directly to theUMTS cell was lower by 80% than traffic offered to the LTEcells.
In the last part of our research we have consideredthe network with traffic overflow between the LTE andthe UMTS systems (Figure 15), under the assumption thatthe LTE cells have different capacities: đ
1= đ2
= 300
allocation units, đ3= 400 allocation units, and đalt = 200
allocation units. The results of the blocking probability forcalls of particular traffic classes are presented in Figure 17.Connections from 3 LTE cells were directed to oneUMTS cell
Mathematical Problems in Engineering 13
V1
V2
V3
LTE
UMTS
Voice Data
Valt
(a)
V1V2 V3
Valt1
......
...
...
A1,1, A2,1, A3,1, A4,1
t1, t2, t3, t4 t1, t2, t3, t4
t1, t2, t3, t4 t1, t2, t3, t4 t1, t2, t3
t1, t2, t3
A1,2, A2,2, A3,2, A4,2 A1,1, A2,1, A3,1
t1, t2, t3, t4
A1,alt , A2,alt ,
A3,alt , A4,alt
(b)
Figure 15: Overflow between cells of the systems: LTE and UMTS, (a) traffic overflow between cells and (b) overflow scheme (đŽđ„,đŠ
is trafficgenerated by calls of class đ„ in cell đŠ and đĄ
đ„is the number of demanded resources by calls of class đ„).
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Bloc
king
pro
babi
lity
Traffic offered [erl]
Simulationâclass 1Calculation (EIG)âclass 1Simulationâclass 2Calculation (EIG)âclass 2Simulationâclass 3Calculation (EIG)âclass 3Simulationâclass 4Calculation (EIG)âclass 4
Simulationâclass 1, altCalculation (EIG)âclass 1, altSimulationâclass 2, altCalculation (EIG)âclass 2, altSimulationâclass 3, altCalculation (EIG)âclass 3, altSimulationâclass 4, altCalculation (EIG)âclass 4, alt
100
10â1
10â2
10â3
10â4
10â5
10â6
10â7
10â8
Figure 16: Blocking probability in mobile system with overflowtraffic: overflow between cells of the LTE system and UMTS.
that serviced additionally its own traffic.The system servicedfour classes of calls, as in the case of the previous consideredcase. The presented results indicate that the differences inavailabilities for particular traffic classes do not influence theaccuracy of the proposed method.
The analysis of the presented results allows us to statethat for each of the considered traffic overflow scheme thepresented method ensures high accuracy of calculations,which is sufficient for practical purposes in engineeringapplications at the stage of analysis and dimensioning ofpresent-day cellular networks.
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Bloc
king
pro
babi
lity
Traffic offered [erl]
Simulationâclass 3 (d2)Calculation (EIG)âclass 3 (d2)Simulationâclass 2, altCalculation (EIG)âclass 2, altSimulationâclass 3, altCalculation (EIG)âclass 3, alt
100
10â1
10â2
10â3
10â4
10â5
10â6
10â7
Simulationâclass 2 (d1)Calculation (EIG)âclass 2 (d1)Simulationâclass 3 (d1)Calculation (EIG)âclass 2 (d1)Simulationâclass 2 (d2)Calculation (EIG)âclass 2 (d2)
Figure 17: Blocking probability in mobile system with overflowtraffic: overflow between cells of the LTE system (different capacitiesof LTE cells) and UMTS; (đ1) is blocking probability in resourcesđ
1
and đ2, (đ2) is blocking probability in resource đ
3.
5. Conclusions
This paper proposes a simple analyticalmethod for a determi-nation of traffic characteristics of mobile systems with trafficoverflow, both single service and multiservice. The proposed
14 Mathematical Problems in Engineering
method is based on the Erlangâs Ideal Grading model withdifferent availabilities and makes it possible to determine theblocking probability in overflow systems based only on theinformation on the capacity of individual network resources(cells) and the value of offered traffic. The proposed methodis characterised by the complexity of orderÎ(đđ2), whereasthe method is based on Fredericks-Hayward method by thecomplexity of order Î(đđ). The higher complexity of theproposed method is compensated by the lack of necessity todetermine the approximate variances.
The paper considers a number of cases that involvedtraffic overflow inmobile networks, that is, both between cells(sectors of cells) of different systems (GSM,UMTS, and LTE),and between cells (sectors of cells) of one system. For eachof the considered overflow scenarios, blocking probabilitiesobtained on the basis of the proposed analytical methodare evaluated. These results are then compared with thedata obtained in the simulations, which confirms very highaccuracy of the proposedmodel.Themodel, farmore simplerthan any earlier developed methods, can be thus used forsolving practical problems in design and optimization ofmodern cellular networks.
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper.
Acknowledgments
Thepresentedwork has been funded by the PolishMinistry ofScience and Higher Education within the status activity taskâStruktura, analiza i projektowanie nowoczesnych systemowkomutacyjnych i sieci telekomunikacyjnychâ in 2015.
References
[1] 3GPP, âUMTS phase 1, release 99,â 3rd Generation PartnershipProject (3GPP) TS 22.100, 2001, http://www.3gpp.org/DynaRe-port/22100.htm.
[2] H. Holma and A. Toskala,WCDMA for UMTS: HSPA Evolutionand LTE, John Wiley & Sons, New York, NY, USA, 2007.
[3] 3GPP, âHigh speed packet access (HSPA) evolution: frequencydivision duplex (FDD) (release 7),â Tech. Rep. 25.999, 3rdGeneration Partnership Project, 2008, http://www.3gpp.org/DynaReport/25999.htm.
[4] âEvolved universal terrestrial radio access (E-UTRA) andevolved universal terrestrial radio access (E-UTRAN); overalldescription; Stage 2 (Release 12) ,â Tech. Rep. 36.300, 3rd Gen-eration Partnership Project (3GPP), 2014, http://www.3gpp.org/DynaReport/36300.htm.
[5] S. Zhao, Z. Xiong, and X. Wang, âOptimal resource allocationfor wireless video over CDMA networks,â IEEE Transactions onMobile Computing, vol. 4, no. 1, pp. 56â67, 2005.
[6] 3GPP, âHandover procedures (Release 11),â 3rdGeneration Part-nership Project (3GPP), TS 23.009, 2012, http://www.3gpp.org/DynaReport/23009.htm.
[7] J. Wang, Q.-A. Zeng, and D. P. Agrawal, âPerformance anal-ysis of a preemptive and priority reservation handoff scheme
for integrated service-based wireless mobile networks,â IEEETransactions onMobile Computing, vol. 2, no. 1, pp. 65â75, 2003.
[8] 3GPP, âHandover requirements between UTRAN and GERANor other radio systems (release 11),â 3rd Generation PartnershipProject (3GPP) TS 22.129, 3GPP, 2012, http://www.3gpp.org/DynaReport/22129.htm.
[9] S. L. Salecker and C. S. Hood, âTraffic analysis of a diversenetwork that overflows dual mode phones,â in Proceedings ofthe IEEE Wireless Communications and Networking Confernce(WCNC â00), vol. 2, pp. 824â828, 2000.
[10] 3GPP, âRadio resource management strategies (Release 7),â3rd Generation Partnership Project (3GPP), TR 25.922, 2007,http://www.3gpp.org/DynaReport/25922.htm.
[11] 3GPP, âGeneral Packet Radio Service (GPRS) enhancements forevolved universal terrestrial radio access network (E-UTRAN)access (release 13),â Tech. Rep. TS 23.401, 3rd Generation Part-nership Project (3GPP), 2014, http://www.3gpp.org/DynaRe-port/23401.htm.
[12] M. Stasiak, M. GĆabowski, A. Wisniewski, and P.Zwierzykowski, Modeling and Dimensioning of MobileNetworks, John Wiley & Sons, 2011.
[13] R. I.Wilkinson, âTheories of toll traffic engineering in theUSA,âBell System Technical Journal, vol. 40, pp. 421â514, 1956.
[14] G. Bretschneider, âDie berechnung von leitungsgruppen furuberfliessenden verkehr in fernsprechwahlanlagen,â Nachrich-tentechnische Zeitung, no. 11, pp. 533â540, 1956.
[15] A. A. Fredericks, âCongestion in blocking systemsâa simpleapproximation technique,â The Bell System Technical Journal,vol. 59, no. 6, pp. 805â828, 1980.
[16] S.-P. Chung and J.-C. Lee, âPerformance analysis and over-flowed traffic characterization inmultiservice hierarchical wire-less networks,â IEEE Transactions on Wireless Communications,vol. 4, no. 3, pp. 904â918, 2005.
[17] M. GĆabowski, K. Kubasik, and M. Stasiak, âModeling ofsystems with overflow multi-rate traffic,â TelecommunicationSystems, vol. 37, no. 1â3, pp. 85â96, 2008.
[18] Q. Huang, K.-T. Ko, and V. B. Iversen, âApproximation ofloss calculation for hierarchical networks with multiserviceoverflows,â IEEE Transactions on Communications, vol. 56, no.3, pp. 466â473, 2008.
[19] E. Brockmeyer, âA survey of A. K. Erlangâsmathematical works,âDanish Academy of Technical Sciences, no. 2, pp. 101â126, 1948.
[20] M. Stasiak, âApproximate model of a switching network car-rying mixture of different multichannel traffic streams,â IEEETransactions on Communications, vol. 41, no. 6, pp. 836â840,1993.
[21] M. Stasiak and S. Hanczewski, âApproximation for multi-service systems with reservation by systems with limited-availability,â in Computer Performance Engineering, N. Thomasand C. Juiz, Eds., vol. 5261 of Lecture Notes in Computer Science,pp. 257â267, Springer, Berlin, Germany, 2008.
[22] M. Glabowski, S. Hanczewski, M. Stasiak, and J. Weissenberg,âModeling Erlangâs Ideal Grading with multirate BPP traffic,âMathematical Problems in Engineering, vol. 2012, Article ID456910, 35 pages, 2012.
[23] H. Akimaru and K. Kawashima, Teletraffic:Theory and Applica-tion, Springer, Berlin , Germany, 1999.
[24] K. Ross,Multiservice Loss Models for Broadband Telecommuni-cation Network, Springer, London, UK, 1995.
[25] U. Herzog and A. Lotze, âDas RDA-Verfahren, ein Streuw-ertverfahren fur unvollkommene Bundel,â Nachrichtentechnis-che Zeitung, vol. 11, pp. 640â646, 1966.
Mathematical Problems in Engineering 15
[26] J. Roberts, V. Mocci, and I. Virtamo, Eds., Broadband NetworkTeletraffic: Final Report of Action COST 242, Commission of theEuropean Communities, Berlin, Germany, 1996.
[27] V. G. Vassilakis, I. D. Moscholios, and M. D. Logothetis, âCall-level performance modelling of elastic and adaptive service-classes with finite population,â IEICE Transactions on Commu-nications, vol. E91-B, no. 1, pp. 151â163, 2008.
[28] J. F. Shortle, âAn equivalent random method with hyper-exponential service,â Performance Evaluation, vol. 57, no. 3, pp.409â422, 2004.
[29] R. Schehrer, âOn the calculation of overflow systems witha finite number of sources and full availiable groups,â IEEETransactions on Communications, vol. 26, no. 1, pp. 75â82, 1978.
[30] G. Bretschneider, âExtension of the equivalent randommethodto smooth traffics,â in Proceedings of the 7th InternationalTeletraffic Congress, Stockholm, Sweden, 1973.
[31] M. GĆabowski, K. Kubasik, and M. Stasiak, âModelling of sys-tems with overflowmulti-rate traffic and finite number of trafficsources,â in Proceedings of the 6th International SymposiumCommunication Systems, Networks and Digital Signal Processing(CSNDSP â08), pp. 196â199, July 2008.
[32] L. Gimpelson, âAnalysis of mixtures of wide- and narrow-bandtraffic,â IEEE Transactions on Communications, vol. 13, no. 3, pp.258â266, 1965.
[33] J. M. Aein, âA multi-user-class, blocked-calls-cleared, demandaccess model,â IEEE Transactions on Communications, vol. 26,no. 3, pp. 378â384, 1978.
[34] J. S. Kaufman, âBlocking in a shared resource environment,âIEEE Transactions on Communications, vol. 29, no. 10, pp. 1474â1481, 1981.
[35] J. Roberts, âA service system with heterogeneous user require-mentsâapplication to multi-service telecommunications sys-tem,â in Proceedings of Performance of Data CommunicationsSystems and their Applications, G. Pujolle, Ed., pp. 423â431,North Holland, Amsterdam, The Netherlands, 1981.
[36] L. E. N.Delbrouck, âOn the steady-state distribution in a servicefacility carrying mixtures of traffic with different peakednessfactors and capacity requirements,â IEEE Transactions on Com-munications, vol. 31, no. 11, pp. 1209â1211, 1983.
[37] E. A. van Doorn and F. J. M. Panken, âBlocking probabilitiesin a loss system with arrivals in geometrically distributedbatches and heterogenous service requirements,â IEEE/ACMTransactions on Networking, vol. 1, no. 6, pp. 664â667, 1993.
[38] J. S. Kaufman and K. M. Rege, âBlocking in a shared resourceenvironment with batched Poisson arrival processes,â Perfor-mance Evaluation, vol. 24, no. 4, pp. 249â263, 1996.
[39] L.-R. Hu and S. S. Rappaport, âPersonal communication sys-tems usingmultiple hierarchical cellular overlays,â IEEE Journalon Selected Areas in Communications, vol. 13, no. 2, pp. 406â415,1995.
[40] X. Lagrange and P. Godlewski, âPerformance of a hierarchicalcellular network with mobility-dependent hand-over strate-gies,â in Proceedings of the IEEE 46th Vehicular TechnologyConference, vol. 3, pp. 1868â1872, May 1996.
[41] M. GĆÄ bowski, âModeling systems with multi-service overflowErlang and Engset traffic streams,â International Journal onAdvances in Telecommunications, vol. 1, no. 1, pp. 14â26, 2008.
[42] Q. Huang, K.-T. Ko, and V. B. Iversen, âAn approximationmethod for multiservice loss performance in hierarchical net-works,â in Managing Traffic Performance in Converged Net-works: Proceedings of the 20th International Teletraffic Congress
(ITC20 â07), Ottawa, Canada, June 17â21, L. Mason, T. Drwiega,and J. Yan, Eds., vol. 4516 of Lecture Notes in Computer Science,pp. 901â912, Springer, Berlin, Germany, 2007.
[43] M. GĆabowski, K. Kubasik, and M. Stasiak, âModeling ofsystems with overflowmulti-rate traffic,â in Proceedings of3rd Advanced International Conference on Telecommunications(AICT â08), p. 6, IEEE, May 2007.
[44] E. Brockmeyer, H. L. Halstrom, and A. Jensen, âThe life andworks of A.K. Erlang,âActa Polytechnika Scandinavia, vol. 6, no.287, 1960.
[45] M. Stasiak and S. Hanczewski, âA model of WCDMA radiointerface with reservation,â in Proceedings of the Interna-tional Symposium on Information Theory and Its Applications(ISITA â08), pp. 1â6, December 2008.
[46] M. GĆabowski, S. Hanczewski, and M. Stasiak, âErlangâs IdealGrading in DiffServ modelling,â in Proceedings of the IEEEAfricon, pp. 1â6, Livingstone, Zambia, September 2011.
[47] A. Lotze, âHistory and development of grading theory,â inProceedings of the 5th International Teletraffic Congress, pp. 148â161, New York, NY, USA, 1967.
[48] M. GĆÄ bowski, A. Kaliszan, andM. Stasiak, âModeling product-form state-dependent systems with BPP traffic,â PerformanceEvaluation, vol. 67, no. 3, pp. 174â197, 2010.
[49] M.GĆabowski, âModelling of state-dependentmultirate systemscarrying BPP traffic,âAnnals of Telecommunications, vol. 63, no.7-8, pp. 393â407, 2008.
[50] M. GĆabowski, M. Stasiak, and J. Weissenberg, âProperties ofrecurrent equations for the full-availability group with BPPtraffic,âMathematical Problems in Engineering, vol. 2012, ArticleID 547909, 17 pages, 2012.
Submit your manuscripts athttp://www.hindawi.com
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttp://www.hindawi.com
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
CombinatoricsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com
Volume 2014 Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Stochastic AnalysisInternational Journal of