End to End (E)GPRS Performance

End-to-End (E)GPRS Performance Improvements BSSG.U. Krause, Com MN PG PLM G1, June 2005

© Siemens 2005 End-to-End GPRS Performance BSS commercially not binding

2

GPRS Performance ImprovementsOverview

In order to evaluate GPRS performances, the main measurements taken in consideration are:

The GPRS improvements shown herein are related to the performance achieved up to BR8.0

To be noted that:• The statements herein depend on the network configuration, the mobiles, and the network elements used. • Additional delay at the terrestrial Abis interface due to e.g. satellite links, Abis compression affects the packet performance

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput Down Link FTP Throughput

Ping Delay

Round Trip Delay

GPRS PDCH activation


3

GPRS Performance ImprovementsGPRS PDCH activation

Down Link FTP Throughput

Ping Delay

Round Trip Delay


Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


4

GPRS Performance Improvements GPRS PDCH activation time

BTSMS BSC

Channel Activation

Time Alignment

Packet Channel Request

First Packet DL Control Block

BR6 BR7

[s]

0.6

0.2

0.4

BR7.0 improvements by• Change the trigger point to start time alignment procedure• Reduction of buffers for each direction in BSC [PPXU only]• Send of the ‘Assignment’ message to MS in advance (parallel processing)BR8.0 improvements by • Optimized evaluation of the Sequence Number to start data transmission

Immediate/Packet Uplink Assignment

Time Alignment

BR8

Optimized improvements with PPXU

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


5

GPRS Performance ImprovementsRound Trip Delay


Ping Delay

Round Trip Delay


Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


6

GPRS Performance Improvements What does mean ‘Round Trip Delay’

MSBSC / PCU

PDAN with Channel Request

PUASProcessing Time

MS-PCU Round Trip Delay (RTD) is intended to be the time interval on the Um interface between

The first Uplink block and the first Packet Uplink Ack/Nack (Contention resolution)

or

The Packet Downlink Ack/Nack (PDAN) with Uplink Request and Packet Uplink Assignment (PUAS)

BTS

RTD

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


7

In order to reduce the MS-PCU Round Trip Delay time

two types of improvement have been provided

1. The reduction of buffers already treated for ‘PDCH Activation Time’ is valid also to reduce the ‘Round Trip Delay’ time

2. Modified internal PCU communication between processor application and HDLC (only SW modifications on the PPXU card)

GPRS Performance Improvements How to achieve reduced ‘Round Trip Delay’

0

250

Round Trip Delay

BR6.0

BR7.0/BR8.0

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput

120200


8

GPRS Performance ImprovementsPing Delay


Ping Delay

Round Trip Delay


Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


9

GPRS Performance Improvements What does mean ‘Ping Delay’

PING DELAY is intended to be the delay between the sending of N bytes IP packet to remote host (query) and the receipt of a confirmation response. (measurements based on series of PING)

The following measurements are based on 50 PING series of 32 byte per PING

BTS BSC SGSNIP server

Ping request in uplink

Ping response in downlink

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


10

GPRS Performance Improvements Configuration & Parameters

The ‘Ping Delay’ time is strongly dependent on the mobile network components that support the command ‘Ping Request’

TEST CONFIGURATION

SIEMENS SGSN

BSS

Default setting of BSS

Coding scheme GPRS CS2 (PCU frame standard)

LAPTOP

WINDOWS 98

TCP receive window and IP packet size optimized

ME 45/S 45 GPRS MS (4+1) or S55 GPRS MS (4+2)

SUN Ultra 5 with SUN OS 5.8 acting as fixed host

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


11

GPRS Performance Improvements Ping Delay Improvements

Apart from the other improvements mentioned before as

Improved buffering structure

Optimized internal communication

additional modifications provide reduced Ping Delay: Modified polling strategy during delayed DL TBF by enlarged fast polling period.

Internal PCU queue is reduced from 3 to 1(or max 2) radio blocks.

Support of the extended uplink TBF mode by open-ended uplink TBF during temporary inactive periods (feature supported by Rel4 MS)

In order to reduce the PING Delay time the BSC supports improvements as follows [PPXU only]:

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


12

GPRS Performance Improvements Ping Delay Improvement Values

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput

* Depends on MS type (might be reduced by 20ms)

** Value with Extended Uplink TBF (valid for MS re. 4 only, e.g. Prototype SIEMENS S75)

Ping Delay in seconds

0

0,2

0,4

0,6

0,8

1

BR6 BR7* BR8**


13

First PING DELAY is intended to be the delay between the sending of the first ping to remote host (query) and the receipt of a confirmation response (measurements based on one PING)

The resource optimized approach by Siemens distinguished between

Channels are already active (1 and 2 phase access)

Channels are idle (1 or 2 phase access)

Note: Idle channels needed to save radio resources (Um, Abis, and PCU) otherwise up to 4 TS per cell at least and the corresponding network resources have to reserved statically, even if no data traffic has to be transmitted

Test configuration is identical to the measurements for Ping series.

GPRS Performance Improvements First Ping Delay

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


14

GPRS Performance Improvements First Ping Delay Improvements

Reduction of sequential PDT/PDCH alignments from three to two. Optimized starting time of the commanded TBF for both active and idle channel

In order to reduce the First PING Delay time the BSC supports improvements as follows [PPXU only]:

Among other improvements mentioned before additional modifications provide reduced First Ping Delay for two-phase access in BR7.0

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


15

GPRS Performance Improvements ‘First Ping Delay’ Values

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput

* Depends on MS type (might be reduced by 20ms)

** Value with Extended Uplink TBF (valid for MS re. 4 only, e.g. Protoype SIEMENS S75)

First Ping Delay in secondsOne Phase Access

0

0,5

1

1,5

2

2,5

BR6 BR7* BR8**

idle channel active channel

First Ping Delay in secondsTwo Phase Access

0

0,5

1

1,5

2

2,5

BR6 BR7* BR8**

idle channel active channel

Competitor Info:First Ping by Ericsson and Motorola is ~650ms (current solution) for active channel, 1 phase access). Idle channel NOT supported (planned for later releases by Ericsson)!


16

GPRS Performance ImprovementsDown Link FTP Throughput


Ping Delay

Round Trip Delay


Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


17

Siemens GPRS DL Throughput Performance

GPRS Performance ImprovementsDown Link FTP Throughput (BR7.0)

Kbit/s

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput

0

50

100

150

200

250

2 TS MS 3 TS MS 4 TS MS

CS 2

CS 4

MCS9


18

GPRS Performance Improvements Down Link FTP Throughput - Configuration & Parameters

The ‘DL FTP Throughput’ time is dependent on the mobile network components.

TEST CONFIGURATION SIEMENS SGSN BSS

Default setting of BSS Coding scheme GPRS CS2 (PCU frame standard)

LAPTOP WINDOWS 98 TCP receive window and IP packet size optimized

SIEMENS S55 GPRS MS (4+2) SUN Ultra 5 with SUN OS 5.8 acting as fixed host

Conditions BERL = 0 (ideal radio condition) FTP download (1000kByte)

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


19

Currently, the focus on data transfer is mainly related to downlink throughput (DL throughput)

Future application like Mobile-to-Mobile data transfers or MMS exchanges requires higher uplink throughput (UL throughput)

GPRS Performance ImprovementsHow to improve GPRS throughput

BR7.0 Uplink balanced assignment of (E)GPRS resources in order

to distribute automatically resources between Uplink and Downlink during running TBF (e.g. TS 4+1 => 3+2)

Dynamic allocation of the number of uplink and downlink time slots (MS multi-slot classes up to 10)

BR8.0 Balanced Uplink and Downlink during running TBF (e.g. TS

4+1, 1+4) and the use of Extended DL TBF Extended dynamic allocation and support of MS multi-slot

classes up to 12

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput


20

GPRS Performance ImprovementsSimulated Gain of DL FTP Throughput BR7.0 vs. BR8.0

Overview

Performance

Improvements

- PDCH activation

- Round Trip Delay

- Ping Delay

- DL FTP Throughput

20

30

40

50

60

70

80

1000 200 100 50 10 5

FTP Download file size [kbyte]

Data

Th

rou

gh

pu

t [k

bit

/s]

BR7 CS2

BR8 CS2

BR7 CS4

BR8 CS4

Simulation Settings:

(4+1) MS, GPRS CS2 or CS4

BR7.0 Rel. 99 vs. BR8.0 Rel. 4/5 (including extended UL TBF)

GR3.1 Parameters, RecWS = 15 * MSS, MSS = 1460 Byte

115 kbps cable modem, Gb-Bandwidth with 2Mbps

Extended UL TBF boosts Throughput of small File Downloads

by up to 20% (MS dependent) !

Extended UL TBF boosts Throughput of small File Downloads

by up to 20% (MS dependent) !


21

Backup Slides [Performance Tables]

Backup Slides


22

COMPARISON of SIEMENS PPCU and PPXU

*) Whole time is less than sum of 3 phases because of parallel Processing, e.g. early Immediate Assignment sending

PPCU/PPXU HW with

BR5.5/6.0

PPXU HW with BR7.0

PPXU HW with BR8.0

GPRS PDCH Channel Activation

80ms 60ms 60ms

Time Alignment Procedure

360ms 140ms 60-80ms

Immediate Assignment

140ms 120ms 120

PDCH bring up Time *)

550 ms 230 ms 150 ms

Backup Slides


23

MS-PCU Round Trip Delay conclusion

120 ms200 ms

MS-PCU Round Trip Delay

PPXU HW withBR7.0/BR8.0

PPCU/PPXU HW with BR5.5/BR6.0

Backup Slides


24

First PING DELAY Table*)

*) Average values; MS type might slightly impact performance

BR6.0 BR7.0 BR8.0

Active Channel, 1 phase access

850 650ms ~600ms

Active Channel, 2 phase access

1050 850ms ~800ms

Idle Channel, 1 phase access

1900ms 1050ms ~800ms

Idle Channel, 2 phase access

2400ms 1150ms - 1250ms

950ms – 1050ms

Backup Slides


25

PING DELAY Table

*) depending on MS type (achieved with NOKIA 6220)

**) BR 8.0 measured with Extended UL TBF (valid only for MS >= Rel.4) depending on MS type (achieved with NOKIA 5140), With fast MS e.g. Prototype Siemens S75

for GPRS (CS2) less than 240ms achievedfor EDGE (MCS6) less than 210 ms achieved

BR5.5 BR6.0 BR7.0*) BR8.0**)

Ping Delay

850ms 720ms 460ms (CS2)

440ms (CS4)

420ms (MCS6)

300ms (CS2)

280ms (CS4)

260ms (MCS6)

Backup Slides


26

Siemens BR 7.0 GPRS DL Throughput Performance

DL Throughput Analysis

(good radio condition, 1 MByte DL)

2 TS MS 3 TS MS 4 TS MS

GPRS CS2 22 kbps 33 kbps ~45 kbps

GPRS CS4 ~26 kbps ~55 kbps ~76 kbps

EGPRS MCS9 ~110 kbps ~164 kbps ~215 kbps

Backup Slides


27

PING BR7.0

Backup Slides


28

FTP Download / GPRS CS4 / MS-Class 10BR7.0

Backup Slides


29

FTP Download / EDGE MCS9 / MS-Class2BR7.0

Backup Slides


30

Measured PING Delay Distribution for 250 successive 32 Byte PINGs

0

20

40

60

80

100

120

200 225 250 275 300 325 350 375 400 425 450 475 500 525 550

PING Delay [ms]

Num

ber o

f PIN

Gs

GPRS CS2 GPRS CS4 EDGE MCS6 GPRS CS2, delayed UL TBF GPRS CS4, delayed UL TBF EDGE MCS6, delayed UL TBF

Gain of the Extended UL TBF

Avg. EDGEPING 260 ms

• Nokia MS 5140 Prototype tested, Rel. 4 compatible, Ext. UL TBF enabled

• With fast Mobile Station e.g. Prototype Siemens S75 less than 210ms achieved

Improvements by Extended UL TBF for Release 4/5 MS BR8.0

Backup Slides


31

PING DELAY (50 Pings GPRS CS2)BR8.0

Backup Slides

237ms in average !

Fast MS Prototype Siemens S75Fast MS Prototype Siemens S75


32

PING DELAY (50 Pings EDGE MCS6)BR8.0

Backup Slides

200ms in average !

Fast MS Prototype Siemens S75Fast MS Prototype Siemens S75


33

NACC in BR8 reduces Cell Reselection Gap BR8.0: Example: FTP download

Backup Slides

Total Gap in Data Transfer is around 600ms to 1200ms depending on MS Type, MS Implementation, UL or DL Traffic (UL GAP is smaller) 1phase or 2phase access Idle / Active Target Cell (PDCH Bring Up / Synchronization required) and Location Area / Routing Area Update required (additional Messages).

Gap from Last Data Block / Packet Cell Change Order in old cell up to Channel Request in new cell is around 260 ms!

Gap from Last Data Block / Packet Cell Change Order in old cell up to Channel Request in new cell is around 260 ms!

New cellOld cell

End to End (E)GPRS Performance

Documents

pcu frame

link ftp throughput

id le ch

parametersoverview

dl ftp throughput

extended ul

round trip

gprs performance