a virtual time simulator for studying qos management functions … · 2016-01-28 · • from the...

1 © NOKIA WCDMA FDD communication / 11.01.2003/ David Soldani for HUT

A Virtual Time Simulator for Studying QoS Management Functions in UTRAN

by David Soldani


Contents

• Simulator structure

• Traffic models

• QoS management functions

• Simulation assumptions

• Simulation results

• Conclusions


GUI

PM

AC

Call generator

LC, PS

Simulator structureGenerate calls/sessions, distribute

UEs and allocate RRP

All calls processed?

Measure DownlinkWide Band Power (PtxTotal)

IfMaxQueueLength is exceeded, then reject call/session and collect statistics

Serve RR based on RRP and arrival time

If time spent in the queue >MaxQueuingTime, then reject the call and collect statistics

If PtxTotal< PtxTarget+Offset & PRT+∆PRT<=PtxTarget, then admit the call/session and collect statistics

Scan queue

NY

Scan queue

Scan queue

If PtxTarget-(PNRT+PRT)>0, or PtxTotal>=PtxTarget+Offset then schedule bit rate based on priorities and arrival times and collect statistics

Scan queue

Show Results

Compute KPIs

Check errors

Move UEs, check for handoffs and update NRT

bit rates

Allcalls/sessions are generated at the beginning and subsequently processed (played back) taking into account the corresponding arrival times and service activities, whence the name virtual time simulator


Traffic models

• Real Time (RT) traffic mix [7]• Conversational Speech (CS), Conversational Data (CD), and Streaming (S) Traffic

Classes (QoS Classes)• Calls are generated according to a Poisson process• For each service a separate process• Each call is held for an exponentially distributed service time• The call inter-arrival period follows the same distribution

• Non-real time (NRT) traffic mix [7]-[8] • Interactive and Background QoS Classes• Arrival of session set-ups is modelled as a Poisson process• For each service there is a separate process• Interactive class

• Each session is modelled as an ON/OFF process• ON periods: log-normal distributed with a cut off• OFF time: Pareto distribution with a cut off

• Each session of Background traffic class is assumed to be in service for an exponentially distributed time and the arrival rate has been approximated with the same type of distribution


Admission Control (1/2)

• Each Resource Request (RR) is assigned a resource request priority (RRP) based on radio access bearer attributes provided by the core network, and a cell based Priority Class parameter, which can be set differently for the different traffic classes

• In case of overload, the radio resource requests are arranged into a queue, and served following the strict priority principle, and, at given priority, taking into account the corresponding arrival times (FIFO)

• Resource requests cannot stay in the queue longer than the maximum allowed queuing time, i.e. Max Queuing Time seconds (TC based management parameter), and are immediately rejected if the maximum allowed queue length, Max Queue Length (cell based management parameter, in number of RABs), is exceeded

• Except for the overload situation NRT traffic is alway admitted


Admission Control (2/2)

• Overload state (1)

• RT traffic is not admitted, if either (1) or the following inequation is satisfied

• The load increase and wide band power estimates are based on thedownlink fractional load equation presented in [5]

OffsetPPPP tTxTargeRTNRTTxTotal +>+=

tTxTargeRTRT PPP >∆+

( )( )DLkkkkkDL ivRWSHO

,11 +−⋅⋅⋅⋅+= αρη


Load (Congestion) Control Function

• The only congestion control actions supported by the simulator is the reduction of the bit rates of NRT bearer services, whenever an overload situation occurs, i.e. when inequation (1) is satisfied

• The bit rates are downgraded starting from the bearer services with lowest priority, and at given priority, based on their arrival times (FIFO), but none of the sessions are released

• Though connections, RT included, may be dropped following the same principle


Packet Scheduler

• Bit rates of admitted NRT bearer services are fast scheduled in a Round Robin fashion (fair throughput), based on resource request priorities and arrival times (FIFO)

• Packet scheduler follows the best effort model and relies on the capacity left by the RT traffic, i.e.

• which will be fairly shared among the different capacity requests whenever is available, whence fast scheduling

• For a fair resources utilisation, the maximum transmission rate (RMax), at given position of the terminal, needs to be computed as a function of the geometry factor (G) and the downlink required transmission power (PRL) for that particular radio link, which yields:

��

��

�

+−⋅=

GPP

NEWR

TxTotal

RL

bMax 11

1

0 α

)(arg RTNRTetTxTNRTAllowed PPPP +−=


Simulation assumptions (1/2)

• Simulation resolution: 500ms

• Simulation time: 6h

• Simulation parameters:Activity factor (ν) 0.67(CS), 1 (CD & Str.) and 0.5 (Background) CCH transmission power 20% Chip rate (W) 3.84 Mcps DCH DL allowed bit rates for Interactive and Background 0, 16, 32, 64, 128, 256 and 384 kbps DCH DL guarantee bit rates for Conversational Data 64 kbps DCH DL guarantee bit rates for Conversational Speech 12.2 kbps DCH DL guarantee bit rates for Streaming 32 kbps (Audio)

64 kbps (Video) Downlink orthogonality (α) 0.5, for ITU Vehicular A Downlink required EbN0 5-7 dB @ 20% FER, depending on bearer service Max Queue length 10 RABs Max Queuing Time 20 s (Em. Call), 10 s (CS & CD), 15s (Str.), and

5s (Int. & Backgr.) Offset 10% Other-to-own cell interference (i) 0.55 Priority Class 1 (Em.Call), 2 (CS), 3 (CD), 4 (Str.), 5 (Int.), and

6 (Backgr.) ARP 1 (Gold), 2 (Silver), and 3 (Bronze) PtxTarget 70 % Simulation time 6 h Soft handover overhead 20%


Simulation assumptions (2/2)

• Radio Resource Priority values:

• Main parameters of simulated traffic:

• ARP and TC values are randomly allocated with equal probability

Call/Session type Mean service time [s]

Mean arrival rate [s]

User traffic [mErl]]

N. of

users Signalling 120 4800 25 10 Emergency call 60 60000 1 5 Conversational Speech 120 4800 25 100 Conversational Data 120 24000 5 80 Streaming 300 10000 30 60 Interactive * * * 70 Background 1200 7200 166 70 * see [8]

Code Call/Session type Gold (ARP=1)

Silver (ARP=2)

Bronze (ARP=3)

1 Signalling 11 11 11 2 Emergency call 1 1 1 3 Conversational Speech 2 2 2 4 Conversational Data 3 3 3 5 Streaming 4 4 4 6 Interactive 5 6 7 7 Background 8 9 10


• 395 users lead to 1125 connection attempts

Call Type

N. o

f Cal

ls o

r Ses

sion

s

0 1 2 3 4 5 6 7 80

50

100

150

200

250

300

350

400

450

Conv.Data

Conv. Speech

Calls/Session arrivals distribution

Signalling

StreamingInteractive

Background

EmergencyCalls

(G)

(S)

(B)

(G)

(S)

(B)


Load status

• All supported QoS management functions work as intended

LRTLNRT PBNRT

LTot


0 10 20 30 40 50 60 70 800

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Lrt(r) Lnrt(g) Ltot(k) [%]

Nor

mal

ised

Dis

tribu

tion

Func

tion

Load distribution

Lnrt Ltot

Lrt


0 200 400 600 800 1000 12000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

RT(r) NRT(g) Total(k) Throughput [kbps ]

Cum

ulat

ive

Dis

tribu

tion

Func

tion

Total traffic

Cell throughput

NRT traffic

NRT traffic


AC queue length distribution

0 0.5 1 1.5 2 2.5 3 3.5 40

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Queue length

DF

(*) a

nd C

DF

(o)


Call Type

Tota

l Cal

l Blo

ck R

atio

[%]

0 1 2 3 4 5 6 7 80

1

2

3

4

5

6

Call Block Ratio

• This demonstrates the impact of prioritising the CS service over the CD service, which in turn has higher priority than the Streaming service, on the selected traffic mix

Conv. Speech

Streaming

Conv. Data


Served RT traffic in Erlangs

• This reflects the corresponding offered loads and call block ratios

Total(1) EmCall(2) ConvS peech(3) ConvData(4) S treaming(5)

Ser

ved

traffi

c [E

rl]

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.50

2

4

6

8

10

12


Interactive and backgroung user throughput

• This demonstrates the impact of prioritising the Interactive service over the Background service

0 20 40 60 80 100 1200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Interactive(o) Background(*) Us er Throughput [kbps ]

Cum

ulat

ive

Dis

tribu

tion

Func

tion


Interactive user throughput

• This demonstrates the impact of prioritising Gold, Silver and Bronze users within the Interactive traffic class

0 20 40 60 80 100 1200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Interactive Gold(r) S ilver(g) Bronze(b) Us er Throughput [kbps ]

Cum

ulat

ive

Dis

tribu

tion

Func

tion


Background user throughput

• This demonstrates the impact of prioritising Gold, Silver and Bronze users within the Background traffic class

0 20 40 60 80 100 1200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Background Gold(r) S ilver(g) Bronze(b) Us er Throughput [kbps ]

Cum

ulat

ive

Dis

tribu

tion

Func

tion


NRT mean user throughput in kbps

• The differentiation among Gold, Silver, and Bronze users is due to the priority based bit rate scheduling, which works as good as expected

NRT Traffic(1) Interactive(2) Background(3)

Av.

Thr

ough

put p

er U

ser [

kbps

]

0.5 1 1.5 2 2.5 3 3.50

5

10

15

20

25

30

35


Conclusions• A virtual time simulator for assessing QoS management functions in UTRAN

has been presented • As a part of this framework, Admission Control, Load Control, and Packet

Scheduler functions, based on priorities and fast scheduling, have been analysed in terms of offered and served traffic mix, throughput, queuing time and call block ratio by means of simulations

• From the simulation results, the proposed solution turns out to perform as intended, and appears to be a good trade of between the complexity and the simplicity of an advanced dynamic and static simulator, respectively; and thus has the potential for investigating any QoS management function in UTRAN, before its deployment throughout a radio access network


References

[1] 3GPP, Technical Specification 23.107, “QoS Concept and Architecture”.

[2] J. Laiho, A. Wacker, "Radio network planning process and methods for WCDMA," Annals of Telecommunications, Vol. 56, No. 5-6, Mai/Juin 2001, pp. 317-331.

[3] Laiho, J., Wacker, A. and Novosad, T. (Editors), “Radio Network Planning and Optimisation for UMTS”, John Wiley & Sons, April 2002, 484p.

[4] S. Hämäläinen, H. Holma, K. Sipilä, "Advanced WCDMA Radio Network Simulator,” Proceedings of PIMRC 1999, Aalborg, Denmark, October 1997, pp. 509-604.

[5] Laiho, J., Wacker, A, Johnson Chris, “Capacity and Coverage Planning for WCDMA” submitted to Wireless Networks, Kluwer Academic publishers’ publication.

[6] Holma, H. & Toskala, A. (Editors), “WCDMA for UMTS”, John Wiley & Sons, April 2000, 344 p.

[7] ETSI, TR 101 112 v.3.2.0, “Selection procedures for the choice of radio transmission technologies of the UMTS”, UMTS 30.03 v.3.2.0.

[8] Shankaranarayanan, N., Jiang, Z. and Mishra, P., “User-Perceived Performance of Web-browsing and Interactive Data in HFC Cable Access Networks”, ICC 2001.

[9] Soldani, D., and Abramowski, M., “An Improved Method for Assessing Packet Data Transfer Across and UMTS Network”, WPMC 2002, Hawaii, October 2002

a virtual time simulator for studying qos management functions … · 2016-01-28 · • from the...

Documents