Вы находитесь на странице: 1из 9

Latency investigation

-including Speedtest.net

Agenda
Objective
Test Cases
Performance Results & Analysis
Prescheduling
Initial UL Buffer Size
SR Periodicity

Moving Forward

Objective
Ericsson has been asked by customer to
investigate the latency time by parameter
changes due to observing poor latency via
Speedtest.net
Areas of testing are:
Prescheduling
Initial UL buffer size
SR periodicity

Test cases
The below test cases were executed while recording TEMs, using a Samsung Galaxy S3
LTE I9305 UE
Basic

32B, 1000B, idle to connected 32B

Connected Mode 100 ping samples 1 second between pings

Optional

Speedtest

Idle Mode

Parameter

Value

Comments

Disabled

default

SR Prediocity

10

default

Initial Buffer Size

86

default

TCBaseline Prescheduling

TC1

Prescheduling
Initial Buffer Size

TC2

Prescheduling
Initial Buffer Size

TC3

TC4

Enabled
86

default

Enabled
300

Prescheduling

Disabled

default

SR Pediocity

10

default

Initial Buffer Size

300

Prescheduling

Disabled

SR Prediocity

Initial Buffer Size

86

default

default

20 samples

Reduce inactivity timer to 10 seconds for purpose for testing

Test are performed on good LTE radio


condition where:
RSRP > -80dBm
RSRQ > -7dB
SINR < 20dB

Comparison
Prescheduling ON and Disabled
Baseline vs TC1

Initial UL buffer size Default vs 300B


Baseline vs TC3, TC1 & TC2

SR periodicity 10 ms vs 5 ms
Baseline vs TC4

Performance Results
Summary

Below is the Average Results of performance

prescheduler = Disabled
prescheduler = Enabled
prescheduler = Enabled, initial buffer size = 300
prescheduer = Disabled, initial buffer size = 300
prescheduler = Disabled, SR periodicity = 5

Performance Results
Summary
Parameter
Prescheduling = Disabled (Default)
TCBaseline (TCB) SR Periodicity = 10 (Default)
Initial Buffer Size = 86 (Default)
Prescheduling = Enabled
TC1
SR Periodicity = 10 (Default)
Initial Buffer Size = 86 (Default)
Prescheduling = Enabled
TC2
SR Periodicity = 10 (Default)
InitialBufferSize = 300
Prescheduling = Disabled (Default
TC3
SR Periodicity = 10 (Default)
Initial Buffer Size = 300
Prescheduling = Disabled (Default)
TC4
SR Periodicity = 5
Initial Buffer Size = 86 (Default)

32B Connected

Delta

29.53

1000B
Connected

Average Latency (ms)


Idle to
Delta
Connected
(32B)

38.07

Delta

248.89

SpeedTest

Delta

36.50

15.60

13.93
(47.18%)

22.94

15.13
(39.74%)

222.33

26.56
(10.67%)

23.20

13.30
(36.44%)

15.61

13.92
(47.13%)

18.98

19.09
(50.15%)

225.44

23.45
(9.42%)

18.10

18.40
(50.41%)

28.13

1.40
(4.72%)

37.46

0.61
(1.61%)

248.28

0.61
(0.25%)

28.70

7.80
(21.37%)

26.47

3.06
(10.36%)

34.06

4.01
(10.54%)

241.89

7.00
(2.81%)

33.00

3.50
(9.59%)

Delta comparison is based on the individual Test Cases (TC) vs Baseline (TCB).
Highest latency improvements observed with prescheduling ON (TC1 & TC2).
Higher initial UL buffer size improves latency for bigger ping sizes, where whole data can be
sent without data fragmentation.
Average of 3ms latency improvement is observed with a lower SR periodicity setting.

PRescheduling

What is causing delays?


Current behaviour for small ping