admission control optimisation and its influence on ... ict-socrates presented by bart sas ibbt fp7...

FP7 ICT-SOCRATES

Presented by Bart Sas IBBT

FP7 SOCRATES final workshop

Karlsruhe, Germany 22 February 2011

Admission Control Optimisation and its

Influence on Handover Optimisation

Contributors: Kathleen Spaey (IBBT), Irina Balan (IBBT), Kristina Zetterberg (EAB), Remco Litjens (TNO ICT)

WWW.FP7-SOCRATES.EU

  Introduction – Admission control – Handover – Goal

 Scenarios – Expected conflicts – Overview – Simulation environment – Approach

 Simulation results  Conclusions

➔ More details can be found in D5.9, section 9.3

Outline

2

Bart Sas, IBBT / University of Antwerp

WWW.FP7-SOCRATES.EU

Introduction

3


WWW.FP7-SOCRATES.EU

  Determines whether a call is admitted to a cell or not –  In order to guarantee QoS

  Distinguishes between fresh and handover (HO) calls –  HO calls are given priority over fresh calls

–  Dropped calls are a greater nuisance for users than rejected fresh calls –  Determined by the ThHO parameter

Admission control (AC)

4


Reserved for HO calls

0 %

100 %

ThHO

reduce rejection fresh

C(t)

reduce rejection HO

WWW.FP7-SOCRATES.EU

  AC parameter optimisation algorithm tunes the ThHO parameter –  In order to adapt the AC algorithm to changes in the environment

Admission control parameter optimisation (AC SON)

5


Collect KPIs: •  Rejection ratio of the fresh calls (RRFC) •  Rejection ratio of the HO calls (RRHOC) •  Low throughput ratio (LTR) •  Traffic loss ratio (TLR)

Optimise: •  If RRHOC, LTR or TLR > Threshold:

lower ThHO •  Otherwise, if RRFC > Threshold: raise

ThHO

WWW.FP7-SOCRATES.EU

  Determines when calls that move away from a cell are handed over to another cell

–  In order to maintain connectivity   The moment when a call is handed over is determined by 2 parameters

–  Hysteresis: the minimum difference between the signal strength of the TenB and the SeNB

–  Time-to-trigger: the amount of time before the handover is triggered

Handover (HO)

6


WWW.FP7-SOCRATES.EU

  HO optimisation algorithm tunes TTT and hysteresis –  In order to adapt the HO algorithm to changes in the environment

  HPI = 2 * CDR + 1 * HOFR + 0.5 * PPHOR

Handover parameter optimisation (HO SON)

7


Collect KPIs: •  Call drop ratio (CDR) •  Handover failure ratio (HOFR) •  Ping pong handover ratio (PPHOR)

Optimise: •  If current HPI > previous HPI:

change optimisation direction •  Adapt Hys/TTT in optimisation

direction

WWW.FP7-SOCRATES.EU

  Admission control (AC) and handover (HO) optimisation algorithms were developed separately

  In reality both algorithms will have to operate in parallel

  In this case both algorithms might influence each other

  Goal of the AC and HO SON integration use case –  Deploy both AC and HO optimisation algorithms together

–  Study interaction between both optimisation algorithms

–  Resolve possible conflicts

Goal of the AC and HO SON integration

8


AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:

Hys, TTT Simultaneous operation:

interaction? Conflicts? Need for integration

AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts

parameters:

ThHO

HO SON Adapts

parameters:



AC SON Adapts parameters:

ThHO

HO SON Adapts parameters:

Hys, TTT

Simultaneous operation: Interaction? Conflicts? Need for integration?

WWW.FP7-SOCRATES.EU

Scenarios

9


WWW.FP7-SOCRATES.EU

  Conflicts between the AC and HO SON algorithms might occur when the AC algorithm rejects a lot of HO calls

–  Because of overload

  Results in high number of call drops –  Because calls do not find a HO target

  HO SON might react –  Because it thinks the HO parameters are wrong –  Reaction is not desired

  Instead, AC SON should react –  By resolving overload

Expected conflicts

10


Overload causes many HO calls to be dropped

HO SON will react

AC SON should react

WWW.FP7-SOCRATES.EU

Scenarios

11


  Evaluate how one of the algorithms affects the targets of the other   Evaluate how the two algorithms work together

  The impact of two different parameters is considered –  UE velocity: influence on HO SON

–  Traffic load: influence on AC SON

# Change Velocity Load

Before After Before After 1 Gradual 3 km/h 50 km/h 2% RRFC 2 Abrupt 3 km/h 50 km/h 2% RRFC 3 Gradual 3 km/h 2% RRFC 20% RRFC 4 Abrupt 3 km/h 2% RRFC 20% RRFC 5 Gradual 3 km/h 50 km/h 2% RRFC 20% RRFC 6 Abrupt 3 km/h 50 km/h 2% RRFC 20% RRFC

A B

A B

WWW.FP7-SOCRATES.EU

  Dynamic system-level simulator

  Downlink direction is simulated

  Network layout: –  25 cells in a 5x5 grid –  500 m site-to-site distance

  Propagation model: –  Okumura-Hata for large urban areas pathloss model –  Both auto- and cross correlated lognormal shadow fading

  Call and traffic generation: –  2500 users generate calls according to a Poisson process –  Real time and non-real time traffic

  Mobility model: –  Random walk

Simulator overview

12


WWW.FP7-SOCRATES.EU

  Scenarios are simulated in four different ways: –  Without any SON algorithm enabled –  With only the AC SON algorithm enabled –  With only the HO SON algorithm enabled –  With both the AC and HO SON algorithms enabled

  Motivation: –  Observe differences between cases –  Deduce influences of the algorithms on each other

Approach

13


AC SON ✗ ✓ ✗ ✓ HO SON ✗ ✗ ✓ ✓

WWW.FP7-SOCRATES.EU

Simulation results

14


WWW.FP7-SOCRATES.EU

0

0.005

0.01

0.015

0.02

0.025

0.03

0.035

0.04

0.045

0.05

0 2000 4000 6000 8000 10000 12000 14000

Rejected Handover Calls

No SONAC SONHO SON

AC & HO SON

Gradual speed and load change − rejection ratio of handover calls

15


RRHOC is lower if AC SON is enabled

Before After

0.00039 0.017

0.00048 0.00078

0.00026 0.023

0.00037 0.00084

WWW.FP7-SOCRATES.EU

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 2000 4000 6000 8000 10000 12000 14000

Rejected Fresh Calls

No SONAC SONHO SON

AC & HO SON

Gradual speed and load change − rejection ratio of fresh calls

16


As a consequence the RRFC is higher

Before After

0.017 0.2

0.05 0.58

0.018 0.24

0.022 0.61

WWW.FP7-SOCRATES.EU

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0 2000 4000 6000 8000 10000 12000 14000

Call Drops

No SONAC SONHO SON

AC & HO SON

Gradual speed and load change − call drop ratio

17


Cases in which AC SON is enabled outperform their counterparts without AC SON

If the HO SON algorithm is enabled the CDR is lower

Before After

0.0045 0.12

0.0057 0.086

0.0098 0.047

0.0081 0.029

WWW.FP7-SOCRATES.EU

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 2000 4000 6000 8000 10000 12000 14000

Ping Pong Handovers

No SONAC SONHO SON

AC & HO SON

Gradual speed and load change − ping pong handover ratio

18


As a trade-off the PPHOR is higher

The lower CDR in case the AC SON is enabled does not cause a higher PPHOR

Before After

0.051 0.22

0.048 0.22

0.057 0.34

0.059 0.33

WWW.FP7-SOCRATES.EU

Conclusions

19


WWW.FP7-SOCRATES.EU

 Scenarios with various changes (speed, load) were defined and run

 The HO SON lowers the CDR at the expense of the PPHOR

 The AC SON lowers the RRHOC at the expense of the RRFC

 The reduced number of rejected handover calls when AC SON is enabled has a positive influence on the CDR

– Calls will be accepted and handed over more rapidly – HO calls will penetrate less far into target cell – Signal quality degrades less – Fewer calls will be dropped – No additional increase of PPHOR

 Expected interaction does not occur –  Because AC SON resolves the problem fast enough –  Interaction might be possible with other implementations

Conclusions

20


WWW.FP7-SOCRATES.EU

Questions?

21


WWW.FP7-SOCRATES.EU

0

0.02

0.04

0.06

0.08

0.1

0.12

0 2000 4000 6000 8000 10000 12000 14000

Call Drops

No SONAC SONHO SON

AC & HO SON

Gradual speed change − call drops

22


Scenarios in which HO SON is enabled

perform better on CDR

WWW.FP7-SOCRATES.EU

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0 2000 4000 6000 8000 10000 12000 14000

Ping Pong Handovers

No SONAC SONHO SON

AC & HO SON

Gradual speed change − ping pong handover ratio

23


As a trade-off the PPHOR is higher

WWW.FP7-SOCRATES.EU

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0 2000 4000 6000 8000 10000 12000 14000


No SONAC SONHO SON

AC & HO SON

Gradual speed change − rejected fresh calls

24


The RRFC is lower because the load is lower due to the higher CDR

WWW.FP7-SOCRATES.EU

 The HO SON lowers the CDR

 As a trade off the PPHOR is higher

 The AC SON does not influence the results

 The same conclusions are valid for abrupt changes

Speed change − summary

25


WWW.FP7-SOCRATES.EU

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0 2000 4000 6000 8000 10000 12000 14000

Rejected Handover Calls

No SONAC SONHO SON

AC & HO SON

Gradual load change − rejection ratio of the handover calls

26


RRHOC is lower if AC SON is enabled

WWW.FP7-SOCRATES.EU

0

0.1

0.2

0.3

0.4

0.5

0.6

0 2000 4000 6000 8000 10000 12000 14000


No SONAC SONHO SON

AC & HO SON

Gradual load change − rejection ratio of the fresh calls

27


As a consequence the RRFC is higher

WWW.FP7-SOCRATES.EU

0

0.005

0.01

0.015

0.02

0.025

0 2000 4000 6000 8000 10000 12000 14000

Call Drops

No SONAC SONHO SON

AC & HO SON

Gradual load change − call drop ratio

28


The CDR is not influenced by the load change

The CDR is not high

WWW.FP7-SOCRATES.EU

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

0.11

0.12

0.13

0 2000 4000 6000 8000 10000 12000 14000

Ping Pong Handovers

No SONAC SONHO SON

AC & HO SON

Gradual load change − ping pong handover ratio

29


The PPHOR is not influenced by the load change

WWW.FP7-SOCRATES.EU

 The AC SON lowers the RRHOC

 As a trade off the RRFC is higher

 The HO SON does not influence the results

 The same conclusions are valid for abrupt changes

Load change − summary

30


admission control optimisation and its influence on ... ict-socrates presented by bart sas ibbt fp7...

Documents