admission control optimisation and its influence on ... ict-socrates presented by bart sas ibbt fp7...
TRANSCRIPT
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
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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:
Hys, TTT Simultaneous operation:
interaction? Conflicts? Need for integration
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:
Hys, TTT Simultaneous operation:
interaction? Conflicts? Need for integration
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:
Hys, TTT Simultaneous operation:
interaction? Conflicts? Need for integration
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:
Hys, TTT Simultaneous operation:
interaction? Conflicts? Need for integration
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:
Hys, TTT Simultaneous operation:
interaction? Conflicts? Need for integration
AC SON Adapts parameters:
ThHO
HO SON Adapts parameters:
Hys, TTT
Simultaneous operation: Interaction? Conflicts? Need for integration?
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
Bart Sas, IBBT / University of Antwerp
Overload causes many HO calls to be dropped
HO SON will react
AC SON should react
WWW.FP7-SOCRATES.EU
Scenarios
11
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
AC SON ✗ ✓ ✗ ✓ HO SON ✗ ✗ ✓ ✓
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Rejected Fresh Calls
No SONAC SONHO SON
AC & HO SON
Gradual speed change − rejected fresh calls
24
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Rejected Fresh Calls
No SONAC SONHO SON
AC & HO SON
Gradual load change − rejection ratio of the fresh calls
27
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
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
Bart Sas, IBBT / University of Antwerp
The PPHOR is not influenced by the load change