MBL3 Sandeep Raina Ver2[1]

1

2

Optimisation and Performance Measurement of GSM/GPRS networks

Telemanagement World ConferenceLas Vegas, 29th-31st October 2002

SANDEEP RAINAGSM & GPRS Consultant

TELSIM,[email protected]

3

Telsim - GSM 900 operator in Turkey

GSM network operations since 1994

First GPRS network in Turkey in 2000

Third GPRS network in World in 2000

GSM world award winner (community category), 2000

41 % market share

Telsim

4

Content

Issues in GSM and GPRS performance data collection Performance Indicators for GSM - Coverage, Capacity, Quality

How Performance Indicators for GPRS are different KPIs for GPRS Benchmarking GPRS KPIs Optimising GPRS - Coverage, Capacity, Quality

5

Issues in collecting performance data

• Lack of standardisation on many interfaces ( PSTN, IN...)• A-bis (Source of Level, Quality and Interference analysis) is customised• A-interface(Source of LU, Call Setup, Handovers) has no details of radio link info. , measurements from BTS or Mobiles

• Multi-vendor environment increases volume of performance data• Matching multi-vendor info. for the simplest analysis• OMC, the main contributor, is a vendor specific device

• Establishing relationship between network elements for a fault (Contribution of each network element in providing fault info.)• Establishing correlation between faults - Symptoms, After-effects

Bad Performance could be due to Inherent / Genetic defect

6

• Validity of extracted data / statistics• Sufficiency of extracted data / statistics• OMC-G does not tap Gb or Gn•Tool unavailable for MS to IP layer monitoring•Drive test is still a snapshot , not the complete picture

•Ascertaining the intensity of problem ; Irregular or no Customer feedback

• How to set a threshold ?…not enough documentation, standards• Hundreds of indicators available...how to make a choice ? Operator experience is not shared• Inflexibility of setting thresholds of operator-interest on Abis, Gn, Gb

• Too many configuration parameters • Missing link between field changes and planning specs.• One-stop controller of configurations not available

Issues in collecting performance data

7

Key Performance Indicators

Set up time

Call drop rateCell

congestion

Cell interference

Handover

RxLevel

Call success rate

The operator concentrates on many varied performance indicatorsThe operator concentrates on many varied performance indicatorsassigning assumed / theoretical priorities...assigning assumed / theoretical priorities...

PRIORITY ASSIGNED TO PERFORMANCE ISSUES : OPERATOR VIEWPOINT

8


Call Drops

Speech Quality

Coverage(Indoor)

No connection

Coverage(outdoor)

Call set up time

...while the subscriber judges the network on the basis of ...while the subscriber judges the network on the basis of a few basic indicators of performancea few basic indicators of performance

9


It is very important to align our KPI definitionIt is very important to align our KPI definitionaccording to according to

what quality and performance means to what quality and performance means to

The SubscriberThe Subscriber

10

Performance Indicators for GSM : Coverage

• Signal Strength - outdoors• In building, In-Car penetration signal levels• Uplink Voice Quality• Downlink Voice Quality• Call Drops • Cell Power control

Start from the gross issues....

11

Performance Indicators for GSM : Capacity

• Erlangs per Cell • TCH success • TCH assignment failures• TCH Drop calls • TCH Blocking • Cell congestion• Congestion Relief Usage• BHCA against rated MSC limit• mErl/subs. against rated MSC limit• SMS/subs. Against MSC limit• MM values(HO,LU,Paging) against limits• Overload : Voice/Signalling/Processor

...to more detailed...

12

• RxLev• Handovers• Call Drops• Call Success Rate• Call set up success rate• Call completion rate• Call set up time • Voice quality(MOS)• RxQual• Echo

• TCH success• TCH assignment failures• TCH Drop calls• SDCCH traffic blocking• SDCCH drop calls• SDCCH Success rates• Location Area Paging• Paging per MSC• PSTN availability and quality

...and finally micro fine-tuning of the network

Performance Indicators for GSM :Quality

13

Why GPRS benchmarking is different

• Multi QoS introduces layered approach to setting benchmarks• Multi handset classes makes the situation a bit more complex• Low Transmission rates on Air Interface• High Radio Link Error on Air Interface

• WWW is low volumes of data and frequent transmissions

• Sequencing of Data is a new concept

• Data Queuing is a new concept

• Blocking probability not easy to define• Different throughput requirements for different applications

• FTP is high data volume and infrequent transmissions

• IP layer performance also influences GPRS performance• TCP layer performance influences GPRS performance

14

How GPRS Benchmarking is different

024681012

h.o.

/re

selec

tion

capa

city

cong

estio

n

call /

sess

ion

erl /

byt

es

inter

fere

nce

cove

rage

time

toco

nnec

t

conc

urre

ncy

traffi

cdis

tribu

tion

GPRS vs GSM - Shift in importance of common indicators

GPRSGSM

Importance level

15

KPI sources

KPI Sources in GSM

KPI Sources in GPRS

Air Interface

Air Interface

Switching network

Core Network

PSTN

IP network

Performance should be measured end to end, element by element

16

KPIs for GPRS

GPRS KPIs can be tapped from :

AIR INTERFACE• Voice Quality• Interference• BLER• RACH Load• UL/DL PDTCH Blocking Rate• Retransmissions on TCP layer ( http, ftp, smtp, pop, wap...)• Network Congestion• Data Compression rate• GPRS contenders per Cell • Cell UL/DL throughput

SESSION STATISTICS• Session Set up success Rate• Session Completion Rate• Dropped packets• Cell Reselections• Processor Loads• Data Traffic transferred• Data blocks erroneously decoded• Data blocks retransmitted

Can be measured

Should be measured

17

CORE COMPLEX• SS7 links congestion• D/L TBF set up success• U/L TBF releases• U/L D/L TBF set up time• Throughput on UL /DL• Data Compression Rates• PDP Context activations • RA updates (subscriber mobility)• Timeslot usage UL and DL

KPIs for GPRS

IP NETWORK• Internet Congestion (Beyond Gi)• Internet Delay (Gi and beyond)• IP traffic per APN• IP traffic on different services• End to end application delays• Round Trip Time• Delays due to Routers and Switches

Can be measured

Should be measured

GPRS KPIs can be tapped from :

18

Tools to collect GPRS KPIs ?

• Collect info. from L2 and L3 layers : RR, MM, MAC, RLC, LLC, GMM, SM, SNDCP• Pick up relevant indicators• Analyse causes and reasons for below-standard performance

• Graphical ,tabular interpretation• Combinations of indicators• Cross-relationship between GPRS indicators• Relationship between GSM and GPRS indicators

C-Value variance vs RLC DL Throughput(non-hopping)

-100

-90

-80

-70

-60

-50

0 10000 20000 30000 40000

Throughput (bps)

Mea

n C-

Valu

e (d

Bm)

19

FTP throughput over Rx level

020406080

100

1 3 5 7 9

11 13 15 17 19 21 23

Test Number

Rx le

vel,

FTP

thro

ughp

ut

Rx level FTP Throuhput

Delay over Rx level

020406080

100

1 3 5 7 9

11 13 15 17 19 21 23

Test number

Rx le

vel,

dela

y in

ds

ec

Rx level Delay in dsec

• Drive Test measurements, RTT measurements

• K1205 Gb protocol tester

Tools to collect GPRS KPIs ?

20

How to use GPRS KPIs ?Some calculations derived from basic KPIs

• What is available ? U/L and D/L bandwidth (CS1 and CS2) available per cell

• What is consumed ? U/L and D/L Channel Loading Factor - Percentage of the available bandwidth in use

Air Interface Data blocks / Total Available air bandwidth GPRS Channel Occupancy - Total data transmit time of GPRS channels of cell/ Total time a channel can be used No. of PDP context activations / subs. - Calculated from GGSN statistics

U/L and D/L Radio Bandwidth - Cell throughput indicator

• What is the frequency of access ? GPRS Access per AGCH - Assignment messages sent on the AGCH of a cell,calculated at OMCR

GPRS Access per PCH - Paging Request messages sent on the PCH of a cell, calculated at OMCR

GPRS Access per RACH - Channel Request messages sent on the RACH of a cell, calculated at OMCR

21

• What is the external influence ? Network IP traffic figure at GTP layer - Directly related to the IP traffic per SGSN/GGSN

Application Overhead Factor - IP Data size in Bytes / IP data actually sent in bytes

Application Throughput - IP Throughput * Application Overhead Factor

• What is the rejection ? PDP activation reject rate - Calculated from stats at GGSN

GPRS Attach reject rate - Calculated from stats at SGSN Total no. of lost packets in the network - Calculated from statistics at SGSN and GGSN

• What is the usage pattern? U/L and D/L RLC Throughput - Total RLC data transmitted / time the GPRS channels were used

Coding Scheme Usage Ratio - Percentage of coding scheme usage per cell (RLC CS1 / RLC CS2)

How to use GPRS KPIs ?Some calculations derived from basic KPIs

22

GPRS Benchmarking

• RxQual (CS2) 0.5 uM• C/I for CS1 9 dB uM• C/I for CS2 12 dB uM • BLER (CS2) 10 % uM• UL PDTCH Blocking Rate < 2% uM• DL PDTCH Blocking Rate <2% uM• TCP Retransmissions <10% uM• RLC/MAC Retransmissions <10% uM• Network Congestion <2% uM• GPRS users per cell(4 ts, CS2) 4 uM• Data Compression Rate 50 % uM• Session set up time 900ms. uM,Core

GPRS BENCHMARK VALUE SOURCE

23

• Session set up success rate 98 % uM,Core (PDP context activation success)• Session drop rate 2 % uM, Core • SS7 link congestion 0.1%Core• DL set up success rate 98 % Core• UL TBF release 2 % Core • UL TBF set-up time 200 ms uM • DL TBF set-up time 400 ms uM • UL throughput (1 ts,CS2) 9 kbps uM• DL throughput (4 ts,CS2) 40 kbps uM• RA updates per subscriber network design value PCU,Core• GPRS pages per subscriber network design value Core• Dropped packets 2% uM,Core,IP


GPRS Benchmarking

24

• Cell reselections Core• Data blocks retransmitted uM• Internet (ISP) Congestion < 3% IP• Internet (ISP) delay < 25 ms(local),<400ms(ext.) IP

• Peak data traffic per subscriber network design value (100 KB/hr.) uM,SGSN• End to end application delay• Round trip latency 900 ms

uM,Core,IP • Router/switch delays < 5 ms Core,IP


GPRS Benchmarking

25

GPRS Optimisation-Capacity

Radio channel capacityGSM channel utilisationGPRS Channel utilisation

Staggered peaks of GSM and GPRS traffic

7:30

TR

AFFI

C

2400 1200 2300 hours GSM + GPRS channel utilisation

Free Channel Capacity

26


23:0

022

:00

Radio channel capacityGSM channel utilisationGPRS Channel utilisation

Concurrent peaks of GSM and GPRS trafficTR

AFFI

C

2400 1200 2300 GSM + GPRS channel utilisation

Free Channel Capacity

27

Accelerator Softwares : Reduction in Data Volume

Application Type

1. Email (SMTP) 2. Text file transfers

3. Zip file transfers 4. JPEGs,GIFs transfers

5. Heavily animated webpages 6. Normal HTML webpages


0

20

40

60

80

100

120

1 2 3 4 5 6

WithoutAccelerationWithAcceleration

Data in

bytes

28

GPRS OptimisationThroughput

Accelerator softwares : Test results

Internet webpage downloads faster • msn.com : 2.2 times faster• nasa.com : 3 times faster• nytimes.com : 3.4 times faster• heavily animated pages : 10 times faster

FTP transfers faster• jpeg file transfer : 2.7 times faster• exe file transfer : 2.7 times faster• zip file transfer : 1.6 times faster

E-mail faster • Uplink Email : 2.3 times faster

29

• Reduce Interference to support higher coding schemes

• Expand GPRS in areas that promise high traffic No need to reserve timeslots across the network for GPRS

• Introduce CS3 and CS4 coding schemes to conserve GSM/GPRS capacity

GPRS OptimisationGeneral

30

2. OSS devices for GSM and specially GPRS need to be more analytic

1. In complex GSM/GPRS networks the role of OSS for management is key

Summary

3. GPRS network management and performance indicators require different treatment and analysis from those of GSM 4. Performance measurement / Optimisation of GPRS is a vital learning exercise for 3G network management

5. QoS is a big question mark for GPRS and therefore for 3G too

31

And what will be the performance standards for

MMS, Audio/Video Streaming formats ?

Current Performance/Optimisation methods provide Best effort services

So, how shall we improve our methods

to provide Guaranteed performance ?

THAT’S THE REAL CHALLENGE NOW !

Challenge

32

SANDEEP RAINAGSM & GPRS Consultant

TELSIMTURKEY

[email protected]

THANK YOU

MBL3 Sandeep Raina Ver2[1]

Documents