Реферат: Evaluating the GPRS Radio Interface for Different Quality of Service Profiles

Abstract.This paper divsents a discrete-eventsimulator for the General Packet Radio Service (GPRS) on the IPlevel. GPRS is a standard on packet data in GSM systems that willbecome commercially available by the end of this year. The simulatorfocuses on the communication over the radio interface, because it isone of the central aspects of GPRS. We study the correlation of GSMandGPRS users by a static and dynamic channel allocation scheme. Incontrast to divvious work, our approach redivsents the mobility ofusers through arrival rates of new GSM and GPRS users as well ashandover rates of GSM and GPRS users from neighboring cells.Furthermore, we consider users with different QoS profiles modeled bya weighted fair queueing scheme. The simulator considers a cellcluster comprising seven hexagonal cells. We provide curves foraverage carried traffic and packet loss probabilities fordifferentchannel allocation schemes and packet priorities as well ascurves for average throughput per GPRS user. A detailed comparisonbetween static and dynamic channel allocation schemes is provided.

1 Introduction

TheGeneral Packet Radio Service (GPRS) isa standard from the EuropeanTelecommunications Standards Institute (ETSI) onpacket data in GSM systems [6], [14]. By adding GPRS functionality tothe existing GSM network, operators can givetheir subscribersresource-efficient wireless access to external Internetprotocol-bases networks, such as the Internet and corporateintranets. The basic idea of GPRS is to provide a packet-switchedbearer service in a GSM network. As imdivssively demonstrated by theInternet, packet-switched networks make more efficient use of theresources for bursty data applications and provide more flexibilityin general. In divvious work, several analytical models have beendeveloped to study data services in a GSM network. Ajmone Marsan etal. studied multimedia services in a GSM network by providing morethan one channel for data services [1]. Boucherie and Litjensdeveloped an analytical model based on Markov chain analysis to studythe performance of GPRS under a given GSM call characteristic [4].For analytical tractability, they assumed exponentially distributedarrival times for packets and exponential packet transfer times,respectively. On the other hand, discrete-event simulation basedstudies of GPRS were conducted. Meyer et al. focused on theperformance of TCP over GPRS under several carrier to interferenceconditions and coding schemes of data [10]. Furthermore, theyprovided a detailed implementation of the GPRS protocol stack [11].Malomsoky et al. developed a simulation based GPRS networkdimensioning tool [9]. Stuckmann et al. studied the correlation ofGSM and GPRS users with the simulator GPRSim [13]. This paperdescribes a discrete-event simulator for GPRS on the IP level. Thesimulator is developed using the simulation package CSIM [12] andconsiders a cellcluster comprising of seven hexagonal cells. Thedivsented performance studies were conducted for the innermost cellof the seven cell cluster. The simulator focuses on the communicationover the radio interface, because this is one of the central aspectsof GPRS. In fact, the air interface mainly determines the performanceof GPRS. We studied the correlation of GSM and GPRS users by a staticand dynamic channel allocation scheme. A first approach of modelingdynamic channel allocation was introduced by Bianchi et al. and isknown as Dynamic Channel Stealing (DCS)[3].

The basic DCS conceptis to temporarily assign the traffic channels dedicated to circuit-switched connections but unused because statistical trafficfluctuations. This can be done at no expense in terms of radioresource, and with no impact on circuitswitched services performanceif the channel allocation to packet-switched services is

permitted only for idletraffic channels, and the stolen channels are immediately releasedwhen requested by the circuit-switched service. In contrast to themodels developed in [4], [9], [10], and [11], our approachadditionally redivsents the mobility of users through arrival ratesof new GSM and GPRS users as well as handover rates of GSM and GPRSusers from neighboring cells. Furthermore, we consider users withdifferent QoS profiles modeled by a weighted fair queueing schemeaccording to [5]. The remainder of the paper is organized as follows.Section 2 describes the basic GPRS network architecture, the radiointerface, and different QoS profiles, which will be considered inthe simulator. In Section 3 we describe the software architecture ofthe GPRS simulator, details about the mobility of GSM and GPRS users,the way we modeled quality of service profiles, and the workloadmodel we used. Results of the simulation studies are divsented inSection 4. We provide curves for average carried traffic and packetloss probabilities for different channel allocation schemes andpacket priorities as well as curves for average throughput per GPRSuser.

3 The Simulation Model

Weconsider a cluster comprising of sever hexadiagonal cells in anintegrated GSM/GPRS network, serving circuit-switched voice andpacket-switched data calls. The performance studies divsented inSection 4 were conducted for the innermost cell of the seven cellcluster. We assume that GSM and GPRS calls arrive in each cellaccording to two mutually independent Poisson processes, with arrivalrates GSMand GPRS,respectively. GSM calls are handled circuit-switched, so that onephysical channel is exclusively dedicated to the corresponding mobilestation. After the arrival of a GPRS call, a GPRSsession begins. During this time a GPRSuser allocates no physical channel exclusively. Instead the radiointerface is scheduled among different GPRS users by the BaseStation Controller (BSC). Every GPRSuser receives packets according to a specified workload model. Theamount of time that a mobile station with an ongoing call remainswithin the area covered by the same BSC is called dwelltime. If the call is still active afterthe dwell time, a handover toward an adjacent cell takes place. Thecall duration isdefined as the amount of time that the call will be active, assumingit completes without being forced to terminate due to handover

failure.We assume the dwell time to be an exponentially distributed randomvariable with mean 1/h,GSMfor GSM calls and 1/h,GPRSfor GPRS calls. The call durations are

alsoexponentially distributed with mean values 1/GSMand 1/GPRSfor GSM and

GPRScalls, respectively. To exactly model the user behavior in the sevencell cluster, we have to consider the handover flow of GSM and GPRSusers from adjacent cells. At the boundary cells of the seven cellcluster, the intensity of the incoming handover flow cannot be

specified in advance.This is due to the handover rate out of a cell depends on the

number of activecustomers within the cell. On the other hand, the handover rate into

the cell depends on thenumber of customers in the neighboring cells. Thus, the

iterative procedureintroduced in [2] is used to balance the incoming and outgoing

handoverrates, assuming that the incoming handover rate hGSM

in i ,

() ( ) −1computed at step i-1.

Since in the end-to-endpath, the wireless link is typically the bottleneck, and given

the anticipated trafficasymmetry, the simulator focuses on resource contention in the

downlink(i.e., the path BSC →BTS→MS)of the radio interface. Because of the anticipated traffic asymmetrythe amount of uplink traffic, e.g. induced by

acknowledgments, isassumed to be negligible. In the study we focus on the radio

interface.The functionality of the GPRS core network is not included. Thearrival

stream of packets ismodeled at the IP layer. Let N be the number of physical channelsavailable in the cell. We evaluate the performance of two types ofradio resource sharing schemes, which specify how the cell capacityis shared by GSM and GPRS users:

thestatic scheme;that is the cell capacity of N physical channels is split into

NGPRS channels reservedfor GPRS data transfer and NGSM = N — NGPRS channels

reserved for GSMcircuit-switched connections.

thedynamic scheme;that is the N physical channels are shared by GSM and

GPRS services, withpriority for GSM calls; given n voice calls, the remaining

N-n channels are fairlyshared by all packets in transfer.

In both schemes, thePDCHs are fairly shared by all packets in transfer up to a

maximum of 8 PDCHs perIP packet («multislot mode») and a maximum of 8 packets

per PDCH [6].

The softwarearchitecture of the simulator follows the network architecture of the

GPRS Network [14]. Toaccurately model the communication over the radio

interface,we include the functionality of a BSC and a BTS. IP packets thatarrive at

the BSC are logicallyorganized in two distinct queues. The transfer queue can hold

upto Q n ⋅8packets that are served according to a processor sharing service

discipline,with n the number of physical channels that are potentially availablefor

data transfer, i.e. n =NGPRS under the static scheme and n = N under the dynamic

scheme. The processorsharing service discipline fairly shares the available channel

capacity over thepackets in the transfer queue. An arriving IP packet that cannotenter

the transfer queueimmediately is held in a first-come first-served (in case of one

priority) access queuethat can store up to K packets. The access queue models the

BSC buffer in the GPRSnetwork. Upon termination of a packet transfer, the IP

packet at the head ofthe access queue is polled into the transfer queue, where it

immediately shares inthe assignment of available PDCHs. For this study, we fix the

modulation and codingscheme to CS-2 [14]. It allows a data transfer rate of 13,4

kbit/sec on one PDCH.Figure 1 depicts the software architecture of the simulator.

Figure 1. SoftwareArchitecture of GSM/GPRS Simulator

To model the differentquality of service profiles GPRS provides, the simulator

implementeda Weighted Fair Queueing (WFQ) strategy.The WFQ scheduling

algorithm can easily beadopted to provide multiple data service classes by assigning

each traffic source aweight determined by its class. The weight controls the amount

of traffic a source maydeliver relative to other active sources during some period of

time. From thescheduling algorithm's point of view, a source is considered to be

active if it has dataqueued at the BSC. For an active packet transfer with weight wi

theportion of the bandwidth i(t)allocated at time t to this transfer should be

() ( ) ⋅∑

where the sum over allactive packet transfers at time t. The overall bandwidth at time

tis denoted by B(t) which is independent of t in the static channelallocation scheme.

Theworkload model used in the GPRS simulator is a Markov-modulatedPoisson

Process(MMPP) [7]. It is used to generate theIP traffic for each individual user in

the system. The MMPPhas been extensively used for modeling arrival processes,

because itqualitatively models the time-varying arrival rate and captures someof the

important correlationsbetween the interarrival times. It is shown to be an accurate

model for Internettraffic which usually shows self-similarity among different time

scales.For our purpose the MMPP is parameterized by the two-statecontinuous-time

Markovchain with infinitesimal generator matrix Qand rate matrix :

The two statesredivsent bursty mode and non-bursty mode of the arrival process.

Theprocess resides in bursty mode for a mean time of 1/andin non-bursty mode for

amean time of 1/respectively.Such an MMPP is characterized by the average

arrivalrate of packets, avgand the degree of burstiness, B.The former is given by:

1 2

Thedegree of burstiness iscomputed by the ratio between the bursty arrival rate and

theaverage arrival rate, i.e., B = 1/avg.

4 Simulation Results

Table 1 summarizes theparameter settings underlying the performance experiments.

Wegroup the parameters into three classes: network model, mobilitymodel, and

traffic model. Theoverall number of physical channels in a cell is set to N = 20

among which at leastone channel is reserved for GPRS. The overall number of GPRS

users that can bemanaged by a cell is set to M = 20. As base value, we assume that

5% of the arrivingcalls correspond to GPRS users and the remaining 95% are GSM

calls. GSM callduration is set to 120 seconds and call dwell time to 60 seconds, so

that users make 1-2handovers on average. For GPRS sessions the average session

duration is set to 5minutes and the dwell time is 120 seconds. Thus, we assume

longer “onlinetimes” and slower movement of GPRS users than for GSM users.The

average arrival rate ofdata is set to 6 Kbit/sec per GPRS user corresponding to 0.73

IP packets per secondof size 1 Kbyte.

Parameter

Figure 2 divsentscurves for carried data traffic and packet loss probabilities due to

buffer overflow in theBSC for the static channel allocation scheme and one packet

priority. For GPRS 1,2, and 4 PDCHs are reserved, respectively. The remaining

channels can be used byGSM calls. With 4 PDCHs the system overloads at an arrival

rate of 0.8 GSM/GPRSusers per second. This corresponds to an average of 12 GPRS

users in the cell (seeFigure 7). In Figure 3 we divsent corresponding curves for the

dynamic channelallocation scheme. For GPRS 1, 2, and 4 PDCHs are reserved,

respectively but morePDCHs can be reserved «on demand». That means that

additional PDCHs can bereserved if they are not used for GSM voice service. From

Figure 3 we observethat for low traffic in the considered cell GPRS makes

effectively use of theon demand PDCHs. For example if 1 PDCH is reserved GPRS

utilizes up to 2 PDCHsat an arrival rate of 0.4 GSM/GPRS users per second. But

with increasing loadthe overall performance of GPRS decreases because of

concurrency among GPRSusers, and more important, priority of GSM users over the

radio interface. Incomparison with the static channel allocation scheme we conclude

that the combination ofreserved PDCHs and on demand PDCH leads to a better

utilization of thescarce radio frequencies. The only advantage of the static channel

allocation scheme isthat it can be realized more easily.

Figure 4 divsents acomparison of overall channel utilization and average

throughput per GPRSuser for the static and dynamic channel allocation scheme. For

the static scheme wereserved 2 and 4 PDCHs respectively and for the dynamic

scheme only 1 PDCH. Weobserve a higher overall utilization of physical channels by

the dynamic scheme.Comparing the dynamic with the static scheme for 2 PDCHs we

detect a slightlyhigher throughput for low traffic load for dynamic channelallocation.

This results from thehigh radio channel capacity available to GPRS users in this case.

They can utilize up to8 PDCHs for their transfer (in contrast to 2 PDCHs in the static

scheme). When loadincreases, GSM calls allocate most of the physical channels.

Thus, throughput forGPRS users decreases very fast. In the static scheme (4 PDCHs)

the decrease inthroughput is not so fast, because GSM calls do not effect the PDCHs.

In an additionalexperiment, we study the performance loss in the GSM voice

service due to theintroduction of GPRS. Figure 5 plots the carried voice traffic and

voice blockingprobability for different numbers of reserved PDCHs. The results are

valid for both channelallocation schemes because of the priority of GSM voice

service over GPRS. Thedivsented curves indicate that the decrease in channel

capacity and, thus, theincrease of the blocking probability of the GSM voice service

is negligible comparedto the benefit of reserving additional PDCHs for GPRS users.

Figure 6 shows carrieddata traffic and packet loss probabilities for the dynamic

channelallocation scheme and different packet priorities. ForGPRS 1 PDCH is

reserved. Weights forpackets with priority 1 (high), 2 (medium), and 3 (low) and

percentagesof GPRS users utilizing these priorities are given in Table 1. Weobserve

that for low traffic inthe considered cell most channels are covered by packets of low

priority. This is dueto the high portion of low priority packets (60%) among all

packets sharing theradio interface. With increasing load medium priority packets and

at last high prioritypackets supdivss packets of lower priority and therefore the

utilization of PDCHsfor low and medium priority packets decreases. For a call arrival

rate of up to 2 callsper second the loss probability of high priority packets is stillless

than 10-5 and thereforethe corresponding curve is omitted in Figure 6.

Figure 7 divsentscurves for average number of GPRS users in the cell and

blocking probabilitiesof GPRS session requests due to reaching the limit of M active

GPRS sessions. Weobserve that for 2% GPRS users the maximum number of 20

active GPRS sessions isnot reached. Therefore, the blocking probability remains very

low. For 10% GPRS usersand increasing call arrival rate, the average number of

sessions approaches itsmaximum. Thus, some GPRS users will be rejected. It is

important to note thatthe curves of Figure 7 can be utilized for determining the

average number of GPRSusers in the cell for a given call arrival rate. In fact, together

with the curves ofFigure 2 and 3, we can provide estimates for the maximum number

of GPRS users that canbe managed by the cell without degradation of quality of

service. For example,for 5% GPRS users and 1 PDCHs reserved, in the static

allocation scheme apacket loss probability of 10-3 can be guarantied until the call

arrival rate exceeds0.4 calls per second, i.e., until there are on the average 6 active

GPRS users in the cell.For the dynamic allocation scheme a packet loss probability of

10-3 can be guarantieduntil the call arrival rate exceeds 0.6 calls per second

corresponding to 9active GPRS users in the cell on average. Figure 8 investigates theimpact of the maximum number of GPRS user per cell to the performanceof GPRS for the dynamic channel allocation scheme with 1 PDCHreserved. Of course, the expected number of GPRS users should be lessthan the maximum number in order to avoid the rejection of new GPRSsessions. On the other hand, the maximum number of active GPRSsessions must be limited for guaranteeing quality of service forevery active GPRS session even under high traffic. The tradeoffbetween increasing performance for allowing more active GPRS sessionsand the

increasing blockingprobability for GPRS users is illustrated by the curves of Figure 8.

Conclusions

This paper divsented adiscrete-event simulator on the IP level for the General Packet RadioService (GPRS). With the simulator, we provided a comdivhensiveperformance study of the radio resource sharing by circuit switchedGSM connections and packet switched GPRS sessions under a static anda dynamic channel allocation

scheme. In the dynamicscheme we assumed a reserved number of physical channels permanentlyallocated to GPRS and the remaining channels to be on-demand channelsthat can be used by GSM voice service and GPRS packets. In the staticscheme no ondemand channels exist. We investigated the impact of thenumber of packet data

channels reserved forGPRS users on the performance of the cellular network. Furthermore,three different QoS profiles modeled by a weighted fair queueingscheme were considered. Comparing both channel allocation schemes, weconcluded that the dynamic scheme is divferable at all. The onlyadvantage of the static scheme lies in its easy implementation. Next,we studied the impact of introducing GPRS on GSM voice service andobserved that the decrease in channel capacity for GSM is negligiblecompared to the benefit of reserving additional packet data channelsfor GPRS. With the curves divsented we provide estimates for themaximum number of GPRS users that can be managed by the cell withoutdegradation of quality of service. Such results give valuable hintsfor network designers on how many packet data channels should beallocated for GPRS and how many GPRS session should be allowed for agiven amount of traffic in order to guarantee appropriate quality ofservice.